RESEARCH DOSSIER
Consumer Surplus and Compensating Variation
Abstract. This monograph examines consumer surplus and Hicksian money-metric welfare measures as instruments for analyzing changes in individual economic opportunities. It derives compensating and equivalent variation from expenditure minimization, explains the relation of ordinary surplus to the Slutsky substitution matrix, establishes conditional rankings and approximation limits, and examines exact compensation along price paths. Reproducible examples demonstrate the consequences of large income effects, binding quantity restrictions, quasilinear preferences, heterogeneous demand and discrete choice. Historical and empirical evidence encompasses household food-price incidence, nonlinear electricity prices, digital platform demand, new product variety, and environmental valuation. The analysis distinguishes the existence of a welfare functional from its empirical identification and from the normative decisions necessary for social aggregation. It develops an original comparative synthesis of established scientific results without advancing a newly named theory or claiming new primary-data estimates.
Keywords: Consumer surplus; compensating variation; equivalent variation; Hicksian demand; Marshallian demand; expenditure function; Slutsky equation; money-metric utility; welfare economics; revealed preference; distributional incidence; demand identification; JEL D11
The object of individual welfare measurement
A change in the price of a commodity alters the set of feasible consumption plans. Consumer welfare measurement asks how much purchasing power, under a specified system of prices and opportunities, is equivalent to that alteration. Its distinctive object is a comparison of attainable living standards for a given decision maker, translated into money through a specified reference situation.
The same movement in a price can generate a change in quantity, a change in expenditure, a change in the value of an existing endowment, and a change in welfare. These magnitudes answer separate questions. An expenditure increase may represent a deterioration in opportunities, an increase in consumption, or both. A fall in expenditure may accompany higher utility when the same product becomes cheaper. A reduction in observed purchases may be the response to a binding affordability constraint rather than evidence of a fall in preferences for the product. The allocation problem connects these observations without identifying them with one another.
The dossier studies consumer surplus, compensating variation (CV), and equivalent variation (EV) as measures of the individual consequences of changes in prices and related opportunities. It extends the analysis to empirical identification, households acting as both buyers and sellers, nonlinear pricing, discrete choice, and nonmarket benefits. The unit of theoretical analysis is an individual or a specified household decision unit. Aggregate macroeconomic stabilization is outside its scope.
The method is deductive and comparative. Standard economic identities are derived from their assumptions, numerical counterexamples are constructed using explicitly hypothetical numbers, and published evidence is interpreted at its stated population and historical horizon. The distinction between theoretical identification and statistical precision governs the assessment. A narrow confidence interval around an incorrectly identified welfare functional has limited scientific value.
Historical development of the surplus concept
In the eighth edition of Principles of Economics, Marshall (1920, Book III, Chapter VI) developed money valuation through the consumer's willingness to give up purchasing power for commodities. Demand provided a practical connection between observed economic choices and the gains associated with access to goods. He also recognized the difficulty of comparing monetary magnitudes across people whose command over resources differed. His presentation belongs to a tradition in which surplus is useful as an approximate measure of individual benefit.
The subsequent ordinal approach sharpened the distinction between a utility level and the expenditure necessary to attain it. Hicks (1943), in The four consumer's surpluses, supplied a historical landmark for distinguishing the price-change and quantity-change problems of welfare measurement. In the modern treatment, price variations hold a reference utility level fixed while quantities are free to adjust. Quantity-constrained problems require a separately specified feasible set. The distinction is especially important when regulated prices coexist with queues or purchase limits.
The practical legitimacy of ordinary surplus received a more precise foundation in Willig's (1976) Consumer's surplus without apology. His approximation analysis tied the magnitude of potential error to income effects and the size of the welfare change relative to available resources. A subsequent correction illustrates why exact welfare calculations need independent numerical checks. Haveman, Gabay, and Andreoni (1987), in a correction to Hausman's example, showed that a corrected labor-tax calculation yielded a 5.2 percent difference between the ordinary and exact welfare-loss measures at a 20 percent tax rate, while the corresponding deadweight-loss discrepancy was much larger (35.5 percent). These are results for the example examined, not general bounds. A small discrepancy in gross welfare loss can become a large proportional discrepancy when tax revenue is subtracted to obtain a comparatively small net loss.
The historical development continued with Hausman's (1981) exact consumer surplus and deadweight loss and Vartia's (1983) compensated-income computation. Their bibliographic records mark research programs in recovering exact welfare measures from ordinary demand. The derivations below are stated independently rather than presented as verified reproductions of every proposition in those articles.
The resulting intellectual structure contains an observable choice relationship, a dual representation of the same optimization problem, and a family of monetary comparisons. It explains why the familiar area under demand remains an indispensable starting point while its welfare interpretation depends on the nature and magnitude of income effects.
Preferences, prices, and the expenditure function
Let x be a vector of nonnegative goods, p a strictly positive vector of prices, m disposable resources in the accounting period, and u(x) a continuous, locally nonsatiated utility representation. Assume an attained maximum on the budget set and an attained expenditure minimum for every utility level under consideration. For the derivative identities, require differentiability and, where necessary, a unique selected demand. At corners the appropriate statements use one-sided derivatives or supporting correspondences.
Ordinary (Marshallian) demand x(p,m) solves utility maximization subject to px <= m. The indirect utility function v(p,m) gives the optimized value. Expenditure e(p,U) is the minimum of px over bundles with u(x) >= U. Compensated (Hicksian) demand h(p,U) solves this second problem. Under the stated duality conditions,
The expenditure function is homogeneous of degree one in prices, concave in prices, and nondecreasing in the required utility level. Multiplying every price by a common positive factor scales minimum expenditure by that factor. Concavity follows because the function is the pointwise infimum of linear expenditure schedules over the utility-contingent feasible bundles. For differentiable e, Shephard's lemma gives e_{p_i}=h_i. An income derivative of v is a conversion factor from money into the locally chosen utility representation. The factor changes under an increasing transformation of utility, whereas an expenditure difference between the same two indifference surfaces survives such a transformation.
Roy's identity follows by differentiating v(p,m) along the expenditure identity: x_i=-v_{p_i}/v_m when v_m is positive. This relationship supplies the local money-metric meaning of a price change. Specifically, a small price increase dp_i at unchanged m costs x_i dp_i to first order. A simultaneous income change dm contributes the corresponding first-order benefit dm. Both statements use the optimized bundle at the point of evaluation.
The two functions answer complementary computational questions. v translates budgets into utility comparisons; e translates utility comparisons into budgets. The compensated income at another price is an output of the expenditure problem, rather than a new arbitrary assumption about tastes. A change in the underlying preference ordering, a rationed allocation, or a transformation in product characteristics requires an explicit enlargement or alteration of the feasible environment.
Compensating and equivalent variation
Write p0 and p1 for the old and new price vectors. Initially hold nominal resources fixed at m. Let U0=v(p0,m) and U1=v(p1,m). Throughout this dossier use gains-positive measures:
CV asks how much could be taken away after the price change while restoring initial utility. EV asks how much would have to be added to the original budget to deliver final utility. For a price increase in a good with positive compensated demand, both measures are negative. The compensation required after the harmful change is -CV. For an improvement, positive CV is a willingness to pay at the new opportunity set; positive EV is the amount under the old opportunity set corresponding to the improvement. Alternative publications use losses-positive signs. A comparison must translate the convention before interpreting a numerical ranking.
When nominal resources change from m0 to m1, define U0=v(p0,m0) and U1=v(p1,m1). The corresponding gains are m1-e(p1,U0) and e(p0,U1)-m0. This form includes cash transfers or an exogenous income shock. Endogenous income arising from factor markets requires calculating the counterfactual resources before applying the comparison.
An equivalent formulation compares expenditure at fixed reference prices: M(p*,U)=e(p*,U). A change M(p*,U1)-M(p*,U0) is a money-metric utility difference. When p*=p0 it equals EV; when p*=p1 it equals CV. The selection of p* determines the units in which welfare is represented. This gives the two measures a common mathematical foundation.
For a single price falling from a to b, with b<a and all other prices fixed,
For increases, the same directed-integral expression is negative. These expressions follow immediately by integrating the price derivative of e along the segment. The baseline utility level belongs to the compensating measure; the final utility level belongs to the equivalent measure. Their possible difference is economically substantive. It quantifies how the value of purchasing-power compensation depends on the reference welfare level.
Suppose the policy is a new entitlement rather than a price reduction. The expenditure function must then carry an environmental state z, and the comparison is between e(p,z0,U) and e(p,z1,U). This generalization connects price welfare to the valuation of clean water, travel reliability, access to public services, or product quality. The EPA economic analysis guidelines (2024, Chapter 7) situate willingness-to-pay and willingness-to-accept questions within the baseline, alternative, and rights governing such comparisons. Those institutions help establish which compensation question should be asked.
Ordinary surplus, Slutsky substitution, and approximation
For a one-good price movement at fixed income, define the gains-positive Marshallian surplus change A as
This integral uses ordinary demand at the same nominal budget throughout the price path. For a normal good over the relevant income and price region and a price reduction, the intermediate income necessary to reach U0 is at most m; the income required to reach U1 at intermediate prices is at least m. Demand nondecreasing in income therefore implies h_i(t,U0)<=x_i(t,m)<=h_i(t,U1). Integrating gives CV<=A<=EV. For a price increase, directed integration reverses the ordering of the integrals and the reference-utility ranking also reverses; with the gains-positive convention the same inequality CV<=A<=EV holds. Under a uniformly inferior good the inequalities reverse. If income effects change sign within the price interval, no single global ranking follows from normality at just one point.
Different measures become identical in a quasilinear environment where the relevant numeraire is consumed in positive quantity and the utility level changes can be translated by unrestricted adjustments of it. Then demand for the affected good is income-independent, compensated demand coincides with ordinary demand, and the three directed areas agree. The qualification about interior numeraire consumption is essential: a binding nonnegativity constraint can reintroduce income effects.
Differentiate the dual demand identity h_i(p,U)=x_i(p,e(p,U)) with respect to price p_j at a fixed U. At a point U=v(p,m),
For smooth preferences with regular demand, the matrix S is symmetric and negative semidefinite because it is the Hessian in prices of expenditure. Ordinary cross-price responses include the income effect and do not generally satisfy symmetry by themselves. Homogeneity and adding-up impose further joint restrictions.
Along a small single-price increase d at the original p and m, write x=x_i(p,m), x_p for its own-price derivative, and x_m for its income derivative. To second order, the positive loss measures are
The difference between exact compensation and ordinary surplus is therefore approximately one-half x times x_m times d^2. This familiar local result also follows by expanding the compensated-income path. For normal goods CV and EV enclose ordinary surplus; the first-order agreement does not require zero income effects. The second-order discrepancy depends on both the magnitude of demand and its sensitivity to resources.
Willig's (1976) approximation bounds provide a finite-change perspective. They are informative because the relevant dimensionless quantities involve expenditure relative to resources and income elasticities. Where these quantities are small and regularity holds over the required region, ordinary demand may suffice for policy precision. Where price changes are large, essential necessities absorb much of the household budget, or preferences are highly income sensitive, exact or bounded compensated measures become more important. The application should report the size of the discrepancy or a defensible bound rather than invoking a general rule of small income effects without checking its relevance.
Compensated paths and multiple price changes
The multicommodity price problem clarifies why exact expenditure differences survive changes in the sequence of price movements. For a piecewise differentiable path p(t) joining p0 to p1 at a fixed utility U, e(p(t),U) has derivative h(p(t),U) dot p'(t). Its integral equals the final expenditure less the initial expenditure. Within a connected regular price region the compensated line integral therefore depends on endpoints. In a differentiable demand representation, symmetry of the Slutsky matrix is an observable implication of this potential-function structure, subject to the information available about income derivatives.
Ordinary demand at fixed nominal income generally is not a gradient field in price space. For two goods i and j its cross derivatives can differ by x_ix_{j,m}-x_jx_{i,m}. A change in the price of one good can alter the demand elasticity of another through a change in real purchasing power. Sequential integration of ordinary surplus over several prices may consequently produce a sequence-dependent total. Adding separate areas computed from independently frozen old demand curves is a different approximation again.
Exact compensation can be recovered without first expressing a utility function algebraically. Starting at m0 and p0, solve E'(t)=x(p(t),E(t)) dot p'(t), with E(0)=m0. Under conditions ensuring an attained solution and uniqueness, the dual demand identity implies E(t)=e(p(t),U0). The negative of E(1)-m0 is CV for unchanged final nominal income. Reverse integration beginning at final income recovers the expense at p0 associated with U1. This is the computational problem associated with Vartia's (1983) methods of measuring welfare change, rather than an assertion that ordinary line integration at fixed income is exact.
For empirical work, errors in estimated demand, discontinuities, integration tolerance, and extrapolation beyond observed price-income support must be recorded. Running an ordinary differential equation solver in two different orders is a useful diagnostic: disagreement beyond numerical tolerance may expose violation of integrability or an inconsistent demand specification. A solver returning a number does not establish that the fitted vector field represents an expenditure function.
Welfare measurement also intersects with index-number theory. Diewert's (1992) exact and superlative welfare indicators investigates connections between welfare comparisons and functions of observed prices and quantities. A cost-of-living index at reference utility U is e(p1,U)/e(p0,U). The same price vector can imply different indexes for households with different preferences and resources. A scalar inflation rate is a convenient descriptive statistic; exact individual compensation for a common reference utility is a more specific object.
Reproducible numerical comparisons
The following calculations are deliberately constructed examples, in units of arbitrary currency. They make no claim about observed consumers. In each case the price unit and period are held fixed.
Cobb-Douglas preferences and a large price shock
Let u(x,y)=x^a y^(1-a), a=0.4, with the price of y equal to one. Let m=100, p0=1 and p1=2. The Marshallian demand for x is am/p and expenditure has the form KUp^a, where K=a^(-a)(1-a)^(-(1-a)). Ratios of expenditures at equal utility depend only on the price ratio r=2. Initial expenditure at new prices is 100r^a and final-utility expenditure at old prices is 100r^(-a).
The positive losses are therefore 100*(2^0.4-1)=31.950791 for compensating variation, 40ln(2)=27.725887 for ordinary surplus, and 100(1-2^(-0.4))=24.214172 for equivalent variation. The approximation falls short of compensation by approximately 4.224904 currency units. It exceeds equivalent loss by approximately 3.511716. The initial expenditure share of x is 40 percent, so the large shock deserves an exact comparison.
The result also shows that affordability after an adverse price change matters. Providing the ordinary-surplus loss as a cash transfer to the new market does not restore the initial utility in this example. The expenditure calculation specifies the additional transfer necessary to reach the reference indifference surface.
A quasilinear benchmark
Let inverse demand for x be P(q)=20-q, q>=0; utility is 20x-x^2/2+y, with y a numeraire and resources sufficient to keep it positive. If price falls from 12 to 8, quantities move from 8 to 12. The ordinary surplus gain is the area of a trapezoid: (12-8)(8+12)/2=40. The consumer can surrender 40 units at new prices and retain the original utility, and would require 40 units at the old prices to become equally well off. The numerical equality is conditional on the absence of a binding numeraire constraint. A sufficiently poor decision maker with identical preferences but an insufficient budget may face a different demand regime.
A corner and rationing counterexample
Consider a consumer whose utility is min(x,y), with both unit prices one and income 10. Without rationing the optimal bundle is (5,5). If purchases of x are capped at two, the highest attainable utility falls to two. Extra nominal income alone cannot restore utility five while the cap persists. The compensating payment required to attain the original utility under the changed constraint is not finite. Ordinary price-based surplus assigns zero to an unchanged sticker price and misses this welfare loss. Feasible-set information is indispensable. The example makes no assumption that real rationing always creates an infinite welfare loss: it identifies a complementary-goods structure in which monetary compensation cannot overcome a physical constraint.
A two-state discrete-choice benchmark
Let the deterministic utilities for an outside option and an inside service be 0 and 1 initially, and 0 and 2 finally. Assume independent type-I extreme-value shocks, a constant marginal utility of income alpha=0.5 per currency unit, and unchanged scale. The expected utility difference divided by alpha is [ln(1+exp(2))-ln(1+exp(1))]/0.5=1.627333. This is an expected money-metric benefit under the specified model, not a guaranteed gain for each realized choice occasion. A different error scale, an omitted outside option, or heterogeneous income coefficients changes its interpretation.
Demand systems, integrability, and revealed preference
The direct recovery of e from a fitted demand system is an empirical implementation of duality. A demand system represents budget shares or quantities across multiple commodities as functions of prices, resources, and observed household characteristics. Deaton and Muellbauer's (1980) Almost Ideal Demand System is a landmark in this tradition. Its budget-share representation begins with a theoretically specified cost function; practical estimation must distinguish exact and approximate price indexes and test relevant restrictions.
A useful demand model satisfies adding-up, homogeneity and Slutsky symmetry in its admissible region, with curvature checked for the fitted expenditure function. A flexible form may fit observed shares while violating monotonicity or curvature outside the sample. Negative predicted quantities, nonpositive expenditure scales, or a positive own compensated price effect signal the need to revise the model or explicitly bound its domain. A global welfare calculation cannot be justified simply by a good in-sample goodness-of-fit statistic.
Revealed-preference methods exploit the price and budget inequalities implicit in observed choices. Afriat's (1967) construction of utility functions is the foundational finite-data approach. Given observations (p_t,x_t), a standard Afriat feasibility formulation seeks numbers U_t and positive lambda_t satisfying, for all s,t,
If the inequalities hold, the minimum of the associated affine upper supports gives a continuous, concave, monotone utility rationalization of these observations under the standard finite-data conditions. This can be checked by substitution at each observed bundle: the t support equals U_t and the remaining supports are at least U_t. For any feasible bundle at t, the t support is at most U_t. Therefore the observed bundle maximizes the constructed representation on the observed budget. This direct reasoning explains the sufficiency of the inequality system.
Varian's (1982) nonparametric demand methods develops the empirical use of such restrictions without requiring a fixed parametric utility form. Rationalizability of a finite dataset establishes existence of at least one admissible preference representation. It does not establish uniqueness, and the set of admissible expenditure functions at prices outside the observed budgets may remain wide. One should report the range of welfare comparisons compatible with the observed choices and maintained assumptions where that range can be computed. A welfare estimate derived from one hand-selected rationalization is a conditional interpolation, not a fully identified empirical result.
Measurement noise creates a further choice. Apparent violations of revealed preference may arise from imperfectly measured prices or bundles, changing household composition, durable goods timing, or actual departures from the model. Relaxing inequalities can diagnose the extent of inconsistency, but the relaxation changes what the resulting welfare bounds mean. It is not a substitute for identifying the origin of the inconsistency.
Econometric identification and the limits of average demand
A demand survey or transaction database records choices at observed prices and incomes. Welfare comparison requires demand or utility at counterfactual prices and, for exact compensation, at counterfactual income. This is an identification problem before it is an estimation problem. A large sample permits precise estimation of conditional means under the observed price assignment; it cannot automatically convert those means into the response to a price intervention.
Price variation may be endogenous to local demand, promotions, product quality, seasonal conditions, or changes in market composition. Researchers can use randomized prices, discontinuities, institutional changes, or instrumental variables when the assignment mechanism and exclusion assumptions are credible. The estimand should describe the population, treatment, outcome, timing, and alternative price regime. A local price response near a cutoff may not characterize the entire demand integral. Instrumental-variable elasticities may be local to people and settings responsive to the instrument. Extrapolation outside the observed range needs independent justification.
Hausman and Newey's (2016) Individual heterogeneity and average welfare identifies a specific difficulty for cross-sectional welfare estimation with general heterogeneous preferences. Observations at a single price and income for each individual identify marginal demand distributions under relevant independence assumptions, but not the complete joint counterfactual demand process for each person. Their analysis motivates bounds based on income effects. This matters because a nonlinear compensation operator applied to a heterogeneous population generally cannot be replaced by the same operator applied to its average demand.
An elementary illustration makes the distinction concrete. Suppose two households have the same m and Cobb-Douglas utility with shares a1 and a2. For a price ratio r, mean compensating loss is m*[ (r^a1+r^a2)/2-1 ]. Computing loss from an average share instead gives m*[r^((a1+a2)/2)-1]. For r not equal to one, the exponential function of a is strictly convex. Jensen's inequality makes the first expression exceed the second for different shares. At a1=0.2, a2=0.6, r=2 and m=100, average exact compensation is 33.220746, while the representative-share calculation gives 31.950791. The approximately 1.269955 difference arises even though demand shares average exactly to 0.4. This standard convexity argument is an illustrative authorial derivation, not an assertion of a newly discovered welfare theorem.
Demand curves can also differ across time horizons. Immediate response measures adjustment using an existing durable stock and established habits; long-run response may include relocation, equipment changes, new competitors, and changes in working time. An estimate of compensation over a day must not silently become an estimate of permanent removal or long-run policy. The correct counterfactual defines what can change and over what interval.
Prices, household surveys, and distributional incidence
The baseline accounting unit must include the complete opportunity affected by the intervention. A price can include a posted charge, tax, shipping obligation, access fee, expected waiting time, or an in-kind requirement. Effective transaction prices can differ across households even when they face the same published price. Unit values (expenditure divided by reported physical quantity) need special care: a higher unit value may reflect the purchase of better quality instead of a higher price for a fixed commodity. Bulk discounts, package sizes and self-produced goods further complicate comparisons.
Household budget surveys identify expenditure patterns and distributional exposures. The official household budget survey methodology describes collection waves and access to anonymized records, while noting remaining cross-country methodological differences, including treatment of owner-occupied housing. A demand system needs corresponding prices, quantities, and resource definitions. An expenditure survey alone cannot recover all household-specific compensated demand derivatives. Survey weights and sampling strata must be used consistently with the target population, and household resources may be adjusted for household size only under an explicit equivalence convention.
For a small price increase of proportional size delta, the first-order money loss for a pure consumer is approximately delta times its original expenditure on the good. Divide by total resources to express this as a fraction of the budget. This calculation is attractive for distributional analysis because baseline shares suffice locally. Large or simultaneous price changes require substitutions, income changes, and relative price interactions to be modeled. Means across income deciles should not replace the within-decile distribution, zero purchases, or uncertainty about actual exposure.
Households that both consume and produce
Consider a household with income from agricultural profit pi(p), other resources m, and consumer demand x. By the envelope theorem for production, a small price increase changes profit by y_i dp_i, where y_i is optimized output. Combining that with the consumer welfare derivative yields a first-order money-metric change (y_i-x_i)dp_i. A household that is a net seller benefits locally while a pure buyer loses. Inputs purchased at affected prices, wages, taxes, and other income effects must be included separately to obtain a broader equilibrium incidence.
Using the 1981-82 Socioeconomic Survey of Thailand, Deaton (1989) analyzed rice prices and income distribution through household consumption and production patterns. His first-round analysis found average rural gains from higher rice prices across living-standard groups, with larger benefits in the middle of the rural distribution. The result is specific to that period, the observed distribution of net rice positions, and the direct incidence considered. It does not establish the consequences of contemporary rice policy in Thailand or other countries, and it does not incorporate all government-revenue or induced-wage effects.
A transferable insight is the accounting relationship, not the numerical or distributional pattern. If a different economy has a high proportion of low-income net buyers, or a different distribution of farm ownership, the same price movement can produce a substantially different distribution of losses. The evidence motivates reporting both sides of the household's exposure and distinguishing static effects from changes in labor markets and fiscal policy.
Nonlinear prices, product variety, and discrete choice
A tariff with multiple blocks, a fixed connection charge or a quantity discount defines a nonlinear budget. The relevant constraint may be y+T(x)<=m, where T is total expenditure rather than a constant unit price times quantity. Changes in T alter both the marginal incentives and the budget available at each quantity. If the feasible set is nonconvex, first-order conditions may not characterize the global optimum. Expenditure minimization can still compare well-defined opportunity sets, but the linear-price Slutsky formulas must not be applied mechanically to a system of tariff parameters.
Ito (2014) used nonlinear electricity pricing across service-area boundaries to study consumers' responses to electricity tariffs. Using monthly administrative household panel data from 1999 to 2007, the study combines a distributional bunching test, tests distinguishing different price measures, and estimation of perceived price. It finds responses consistent with average rather than marginal or expected marginal prices in the Southern California setting studied. The policy implications depend on the behavioral mechanism, and a welfare calculation assuming full marginal-price optimization can therefore be sensitive to that maintained model. A consumer responding to an average bill may face different perceived incentives from a consumer who knows the full schedule.
When choice involves indivisible products, the welfare object is often a change in the maximum of choice-specific utilities. Small and Rosen (1981), in discrete-choice welfare economics, are a historical reference for applied welfare methods in this domain. With deterministic choice values V_j, independent identically distributed type-I extreme-value shocks of unit scale, and a common constant positive money coefficient alpha, expected maximum utility equals a constant plus log(sum exp(V_j)). Consequently, dividing the difference between two logsum terms by alpha gives the expected benefit in currency units, provided the utility scales and choice environments are comparable.
A wider choice set can raise attainable utility while reducing the share of an existing product. Demand for a single seller's product is therefore not an adequate measure of the entire gain from entry or the loss from exit. Product elimination changes the set of substitute options, and the relevant counterfactual may involve a price high enough to eliminate demand or an explicit removal from the choice set. The outside option must have a meaningful interpretation. A binary purchase model that treats all nonpurchases as equivalent may conceal substitution to rival suppliers, waiting, or home production.
Feenstra's (1994) research on new product varieties connects changes in available varieties to import-price measurement. The implication for welfare analysis is that comparison of prices for continuing items alone omits an opportunity-set margin. The magnitude of the omitted gain depends upon the substitution possibilities and valuation of the new products. This dossier does not infer national gains from trade or compare aggregate stabilization policies; it identifies the individual choice margin relevant to such subsequent studies.
Empirical evidence from digital services and nonmarket goods
Transaction-level data can reveal differences in choices under similar observed conditions. Cohen, Hahn, Hall, Levitt, and Metcalfe (2016), in Using big data to estimate consumer surplus, studied UberX in the United States using price variation. The institutional University of Chicago research account identifies the rounding of surge prices as a source of variation and records almost 50 million customer sessions in four US cities. It also reports the study's extrapolated 2015 US annual consumer surplus of approximately $6.8 billion. This is a reported working-paper estimate, not a new result calculated from underlying transactions here.
The economic design offers both information and limits. Discontinuities in displayed prices can reveal local responses, yet making an integral over demand requires assumptions about responses away from the observed thresholds. A short-run loss associated with removing a service for a day differs from a permanent removal during which consumers and suppliers reorganize. Moreover, consumer benefit is one component of social welfare; effects on drivers, platform profit, competitors, congestion, taxes and other external effects require separate accounting. The 2016 study's number should not be converted into a current estimate or a universal platform valuation.
Digital goods with zero posted monetary prices sharpen the difference between observed expenditure and benefit. A service may be valuable because of social interaction, information or entertainment even when no money changes hands at the point of consumption. Time, data disclosure, advertising exposure, and forgone alternatives still have potential opportunity costs. Brynjolfsson, Collis, and Eggers (2019), in massive online choice experiments, examine ways to elicit welfare valuations for such goods through choices involving monetary compensation or access. The work provides evidence on a measurement approach rather than a universal monetary value of free services.
An experiment paying participants to surrender access measures a particular relinquishment question for a specified duration. Its monetary value depends on who is eligible, what substitutes are available, whether compliance is monitored, and whether the compensation is real or hypothetical. Assigning that value to all users in a country or across an indefinite time horizon requires evidence that the relevant valuation remains stable. Survey nonresponse and self-selection can change the distribution of measured willingness to accept. Reporting the median is a descriptive choice and does not recover the sum of individual gains.
Environmental and other nonmarket valuation poses a related challenge. The EPA valuation discussion distinguishes monetary tradeoffs for improvements and losses and situates preference-based valuations in benefit-cost analysis. A stated-preference experiment needs a comprehensible counterfactual, credible payment vehicle, treatment of uncertainty, and a clear description of the affected entitlement. Revealed-preference approaches can value observable behavior associated with a nonmarket attribute, subject to assumptions about complementary markets and choice. Neither method by itself turns a consumer valuation into an exhaustive social or ethical assessment of the underlying environmental outcome.
Income heterogeneity and welfare aggregation
Each individual compensation amount is expressed in currency units at a specified price reference. Summing these amounts is a useful accounting operation, provided there is a common time interval, currency, relevant population, and direction of welfare change. It gives an unweighted monetary measure and an indication of the scale of potential compensation. It is not a direct addition of interpersonal utilities.
If a social evaluation function is differentiable in individual utilities and each utility is increasing in income locally, its marginal response to a transfer has a welfare weight proportional to the product of the social derivative with respect to individual utility and that individual's marginal utility of money. Computing a weighted sum requires choices about the social objective and the interpretation of its parameters. A policy with positive total monetary surplus can impose substantial harm on particular households. The weights and the incidence of gains and losses are distinct disclosures.
For finite movements, even a fixed weight per currency unit embeds a reference-price convention. A weighted sum of equivalent variations can differ from an otherwise identical weighted sum of compensating variations. The analyst should choose the money metric appropriate to the policy question and avoid treating a disagreement among these totals as a mere rounding issue. Alternatives can be evaluated by presenting the distribution of both measures when feasible.
Imagine households A and B experience monetary effects of +50 and -40. The unweighted total is +10, suggesting a feasible hypothetical transfer under simplified zero-cost transfer assumptions. If the loss falls on a household with twice the social weight of the gaining household, the illustrative weighted sum is 50-80=-30. These values merely illustrate the dependence of an evaluated ranking on explicit weights; they do not identify which weights a government or society should adopt. Furthermore, a transfer actually paid may alter behavioral incentives, administrative costs, and the resulting equilibrium. Potential compensation is not a record of completed compensation.
The distinction between consumer welfare and social welfare is critical for regulated products and taxation. Government revenue and supplier surplus may change simultaneously. A payment from a buyer to a seller is a loss to the buyer and receipt to the seller; it is not necessarily a net resource loss. Real production costs, administrative costs, external damages and benefits, and distributional consequences require their own estimates. A partial-equilibrium consumer calculation can inform a broader assessment, but the broader assessment must explicitly construct the relevant additional accounts.
Policy counterfactuals, taxes, and cost of living
Suppose a competitive market has a tax-free supply price c, constant real marginal cost, and quasilinear consumer demand Q. A unit tax t raises the consumer price from c to c+t. The consumer-surplus loss is the integral of Q over that price interval. Government collects t*Q(c+t). Under the simplified assumptions of no externalities, unchanged producer surplus, and public revenue valued at one currency unit per unit received, the remaining loss is the integral of Q(p)-Q(c+t) from c to c+t. It is the value of transactions forgone relative to the tax payment transferred to government. The familiar triangular approximation relies on local linearity of demand.
When consumers have substantial income effects, exact money-metric losses and the value of revenue cannot always be combined as if the quasilinear assumptions held. A price subsidy may reduce consumer losses but induce public expenditure, tax-financing distortions, or changes in consumption and supply. The transfer objective and the resource-cost objective deserve separate comparisons. Finkelstein and Hendren (2020), in Welfare analysis meets causal inference, discuss the marginal value of public funds as a way to relate causal policy evidence to welfare measurement. Its policy scope is broader than the individual CV functional: the value of benefits and the net governmental cost must both be identified.
Cost-of-living calculations ask how much expenditure is necessary to maintain a reference standard of living across price systems. A fixed basket price index records the cost of purchasing a specified bundle. A fully optimizing consumer can substitute among feasible alternatives; the expenditure index measures the minimum cost of meeting the reference utility. When prices and resource distributions change substantially, reporting a single index for households with different budget shares obscures heterogeneity in compensation needs. Some households may also lack access to substitutes available to the average consumer.
The policy counterfactual must specify eligibility, geography, timing and financing. If a price ceiling reduces the posted price but restricts quantity through queues or shortages, evaluating welfare using only the price change assumes away a central feature of the intervention. If a tax or transfer changes nominal income as well as prices, U1 must be computed with the changed budget. If a quality improvement associated with a higher price is mandatory, comparisons holding quality fixed address an artificial intervention. All three cases show why consumer welfare calculation is a problem about opportunity sets, not merely observed expenditure changes.
Scientific measurement, robustness, and uncertainty
The analytical process begins with the identification of the object: a change in price, income, quality, access or rationing, and the reference state in which monetary compensation is evaluated. Specify which goods and time interval are covered and whether observed households can adapt their durable stocks, work hours or production decisions. Report whether the target is an individual, a household, a selected user population or a weighted population.
Next assemble the evidence required by the chosen estimator. A local first-order calculation needs baseline quantities and price changes. Ordinary surplus needs a demand curve over the relevant interval. Exact compensation needs an integrable demand system, an expenditure function, a compensated-income path, or sufficiently informative revealed-preference bounds. A discrete-choice method needs alternatives, comparable choice utilities, a scale and money coefficient, and support for the counterfactual change. Where evidence identifies only bounds, presenting a uniquely determined amount implies precision beyond what the assumptions support.
Uncertainty has several origins. Sampling uncertainty concerns finite observations and survey design. Parameter uncertainty concerns a fitted model and its estimated covariance. Identification uncertainty concerns multiple functional relationships consistent with observations. Scenario uncertainty concerns the relevant future price, availability and income responses. Model uncertainty concerns alternative admissible preference and behavioral specifications. These should be distinguished because they call for different remedies: a bootstrap can approximate sampling variability but cannot repair an unidentified counterfactual.
Useful checks include reproducing expenditure identities at observed prices, testing budget adding-up, comparing Marshallian and compensated estimates, verifying price-path invariance under the maintained model, and repeating calculations under alternative demand specifications. Report parameter uncertainty in the nonlinear welfare functional rather than transferring standard errors from price elasticities mechanically. For distributional estimates, preserve household sampling weights and clustering when resampling, and examine sensitivity to outliers and sparse expenditure categories.
For ordinary surplus, sensitivity to income effects is particularly important. A credible interval for income elasticity combined with the scale of the price movement can guide an approximation analysis. For nonmarket valuation, examine treatment wording, scope, entitlement, realistic budget constraints, and follow-up compliance where applicable. For platform data, distinguish access to a firm-specific dataset from representative population sampling and report the provenance of each estimate.
The literature in this dossier is examined at different access levels: directly available analytical texts and official methodology, and bibliographic or abstract-level records for restricted papers. Historical attributions based only on records are confined to the research topic and publication identity. Empirical claims are limited to the reported designs and results present in the accessible records. No access to proprietary transactions, protected household microdata or complete statistical replication is claimed. The examples and formal propositions are internal deductive calculations independent of inaccessible articles.
Analytical synthesis and research implications
The preceding results permit a unified assessment of when ordinary surplus is sufficient for individual consumer welfare work. Ordinary surplus is exact when the affected demand is independent of income along the relevant path and the corresponding compensation is feasible. It is a first-order measure for smooth price changes in the standard model. It can be a close finite-change approximation when income effects and the relative scale of the movement support that conclusion. It can become inadequate in substantial affordability shocks, heterogeneous populations, multiple nonintegrable fixed-income price paths, and environments where actual access or quantity constraints change.
Three analytically distinct requirements emerge. The reference state determines which exact monetary comparison is relevant. The opportunity set determines what the consumer can attain and what compensation can restore. The empirical design determines which parts of counterfactual demand and income responses are identified. Keeping these requirements explicit prevents an observed behavior, an abstract theoretical object and a policy judgment from collapsing into a single quantity.
The comparative cases reveal a common structural problem. A household exposed to a rise in the price of rice may simultaneously receive an increased selling price. A household facing nonlinear electricity bills may optimize with respect to a perceived price that differs from the tariff's marginal rate. A user of a digital service may face the loss of an available option without ever paying a positive posted price. What transfers across these cases is the obligation to define the counterfactual opportunity and the associated money metric. The numerical results do not transfer without evidence.
An important authorial inference is that robustness assessment should target the largest approximation error relevant to the substantive question rather than demand maximum modeling complexity in every application. When a small conventional price movement affects a minor budget item, a well-identified ordinary demand response may yield more useful evidence than a weakly identified full structural model. When the same movement threatens a basic household requirement or changes feasible alternatives, the exact comparison or a scientifically credible bound should take priority. This is an application of established theory and identification logic rather than a new named model or doctrine.
Another implication concerns communication. A distributional table should separate money gains, money losses, resource changes and weights, and show the reference-price convention. Such reporting lets a reader reconstruct the economic comparison before considering policy judgments. A consumer-surplus figure with no clearly defined counterfactual and population is incomplete even when its arithmetic is correct. A carefully bounded welfare comparison can be more consequential than a very precise-looking point estimate that hides its identifying assumptions.
Future research can improve the joint observation of income and price responses for heterogeneous households, develop credible bounds where only partial demand information is available, measure substitution under rationing and product entry, and integrate individual choice with empirically justified models of perceived prices. Data sharing, documented survey design and reproducible calculation code can reduce uncertainty about empirical implementation. These priorities follow from the demonstrated limits of available observations and established welfare methods; no new theoretical system or exclusive claim of scientific novelty is advanced.
Conclusion
Consumer surplus and compensating variation concern how changes in economic opportunities translate into monetary comparisons of individual attainable welfare. The expenditure function supplies the exact money metric, while compensated demand gives its price derivative. Equivalent variation uses the original price system as the reference; compensating variation uses the final price system. Ordinary consumer surplus integrates fixed-income demand and agrees with the exact measures to first order under smooth regularity, becoming exactly equal when income effects are absent on the relevant feasible path.
For large or heterogeneous changes, the difference among these measures has an economic meaning. It expresses the role of purchasing power in shaping substitution and the compensation necessary to reach a specified living standard. Multi-price comparisons require integrability or appropriately bounded counterfactual demand. Nonlinear tariffs, quantity controls, dual producer-consumer roles and new products require the actual opportunity set and resource changes to enter the analysis.
The empirical literature shows both the possibilities and limits of recovery from household surveys, quasi-experimental price variation, platform transactions and stated choices. Identification of the relevant demand relationship, comparability of monetary units and transparent scope precede statistical estimation. Individual valuations can inform distributional and social evaluation, while aggregation, transfers, real resource costs and explicit welfare weights remain separate tasks. The enduring scientific value of the surplus concept is its capacity to connect observable economic choices to a clearly defined welfare comparison, under conditions that can be stated, examined and challenged.
References
Afriat, Sydney N. (1967). The construction of utility functions from expenditure data. International Economic Review, 8(1), 67-77. doi:10.2307/2525382.
Brynjolfsson, Erik, Collis, Avinash, & Eggers, Felix. (2019). Using massive online choice experiments to measure changes in well-being. Proceedings of the National Academy of Sciences, 116(15), 7250-7255. doi:10.1073/pnas.1815663116.
Cohen, Peter, Hahn, Robert, Hall, Jonathan, Levitt, Steven D., & Metcalfe, Robert. (2016). Using big data to estimate consumer surplus: The case of Uber. NBER Working Paper 22627. doi:10.3386/w22627.
Deaton, Angus, & Muellbauer, John. (1980). An almost ideal demand system. American Economic Review, 70(3), 312-326.
Deaton, Angus. (1989). Rice prices and income distribution in Thailand: A non-parametric analysis. Economic Journal, 99(395), 1-37. doi:10.2307/2234068.
Diewert, W. Erwin. (1992). Exact and superlative welfare change indicators. Economic Inquiry, 30(4), 565-582. doi:10.1111/j.1465-7295.1992.tb01282.x.
Eurostat. (n.d.). Information on data: Household budget surveys. European Commission. Retrieved November 17, 2026.
Feenstra, Robert C. (1994). New product varieties and the measurement of international prices. American Economic Review, 84(1), 157-177. Working paper antecedent: New Goods and Index Numbers: U.S. Import Prices, NBER 3610 (1991). doi:10.3386/w3610 (working paper).
Finkelstein, Amy, & Hendren, Nathaniel. (2020). Welfare analysis meets causal inference. Journal of Economic Perspectives, 34(4), 146-167. Working paper: NBER 27640. doi:10.3386/w27640 (working paper).
Hausman, Jerry A. (1981). Exact consumer's surplus and deadweight loss. American Economic Review, 71(4), 662-676.
Hausman, Jerry A., & Newey, Whitney K. (2016). Individual heterogeneity and average welfare. Econometrica, 84(3), 1225-1248. doi:10.3982/ECTA11899.
Haveman, Robert H., Gabay, Mary, & Andreoni, James. (1987). Exact consumer's surplus and deadweight loss: A correction. American Economic Review, 77(3), 494-495.
Hicks, John R. (1943). The four consumer's surpluses. Review of Economic Studies, 11(1), 31-41. doi:10.2307/2967517.
Ito, Koichiro. (2014). Do consumers respond to marginal or average price? Evidence from nonlinear electricity pricing. American Economic Review, 104(2), 537-563. doi:10.1257/aer.104.2.537. Working paper version: NBER 18533 (2012).
Marshall, Alfred. (1920). Principles of Economics (8th ed.). London: Macmillan. Book III, Chapter VI, Value and Utility.
Small, Kenneth A., & Rosen, Harvey S. (1981). Applied welfare economics with discrete choice models. Econometrica, 49(1), 105-130. Working paper version: NBER 319 (1979). doi:10.3386/w0319 (working paper).
U.S. Environmental Protection Agency. (2024). Guidelines for Preparing Economic Analyses (3rd ed.). EPA-240-R-24-001. December 2024.
University of Chicago. (2016). Big data gives insight into appeal of services like Uber. UChicago News.
Varian, Hal R. (1982). Non-parametric methods in demand analysis. Economics Letters, 9(1), 23-29. doi:10.1016/0165-1765(82)90092-1.
Vartia, Yrjo O. (1983). Efficient methods of measuring welfare change and compensated income in terms of ordinary demand functions. Econometrica, 51(1), 79-98. doi:10.2307/1912249.
Willig, Robert D. (1976). Consumer's surplus without apology. American Economic Review, 66(4), 589-597.