RESEARCH DOSSIER
Price and Income Elasticities of Demand
Abstract. Demand elasticity measures the proportional response of a specified quantity demanded to a specified economic change. Its scientific interpretation depends on the definition of the commodity, the decision maker, the budget, the price vector, the adjustment horizon, and the counterfactual under examination. This monograph reconstructs the historical foundations, mathematical restrictions, empirical identification, and economic applications of own-price, cross-price, and income elasticity. It derives the elasticity form of the Slutsky equation, identifies the implications of adding up, homogeneity, symmetry, and compensated negativity, and demonstrates the compatibility of inferior and Giffen demand with constrained utility maximization. Engel curves, subsistence expenditure, flexible demand systems, heterogeneous consumers, discrete purchase, quality adjustment, and durable goods explain why observed responses vary within and across populations. The empirical analysis examines revealed-preference tests, household surveys, randomized price variation, differentiated-product estimation, and international expenditure comparisons. Evidence from China, Côte d’Ivoire, Tanzania, Ethiopia, Mexico, the United Kingdom, the United States, the European Union, medical insurance, gasoline, and global purchasing-power comparisons is integrated within its distinct measurement and identification boundaries. An original analytical synthesis distinguishes individual compensated responses, household budget responses, population aggregation, and equilibrium price-quantity movements. The monograph establishes a practical basis for evaluating elasticity estimates, conducting policy analysis, and transferring empirical findings while preserving the assumptions under which each result has meaning. Mathematical examples are constructed for exposition; the work makes no claim to new empirical estimates or a newly established theory.
Keywords: Demand elasticity; own-price elasticity; cross-price elasticity; income elasticity; expenditure elasticity; Marshallian demand; Hicksian demand; Slutsky equation; Engel curve; Giffen goods; demand systems; AIDS; QUAIDS; revealed preference; identification; unit values; aggregation; tax incidence; consumer welfare; microeconomics
The scientific object of elasticity
A demand function relates a quantity selected by an economic decision maker to the prices, resources, information, and other constraints governing that choice. An elasticity expresses one local response of this function in proportional terms. The object is therefore a derivative of a behavioral relation evaluated at a particular point, or a specified approximation to that derivative over a finite change. An elasticity belongs to a function together with its arguments and the conditions defining their variation.
The distinction is consequential because a single commodity may support several legitimate elasticity estimates. Weekly retail purchases of gasoline respond to pump prices under an existing distribution of vehicles. Annual gasoline use responds additionally to changes in journeys, residential arrangements, vehicle efficiency, and the selection of vehicles. Household fuel expenditures may rise even as physical consumption falls. A retailer's own-brand sales can respond more strongly to its own price than total consumption responds to the market-wide price of the category. Each object has its own economic boundary and its own counterfactual.
A scientific study defines the physical or service quantity, the unit of observation, the monetary resource, the relevant price vector, the adjustment interval, and the population before interpreting a derivative. An observed decline in sales after a price increase is an event. A causal demand elasticity is a property of a response under a counterfactual change in the price, with the remaining conditions specified. Their coincidence is a matter for an identification design to establish. The structure of the remainder of this monograph follows from this distinction between the economic response to be measured and the conditions under which data reveal it.
Historical development of the demand response
The nineteenth-century development of marginal analysis made it possible to organize demand around incremental choices. In Marshall’s analysis of elasticity (1890), proportional change gave responsiveness a common scale across otherwise incomparable commodities and locations on a demand schedule. Marshall's treatment also directs attention to variation in responsiveness along that schedule. A single numerical coefficient is consequently a special representation, not the definition of demand behavior itself.
The next step was the separation of the changes produced by a price movement. Slutsky’s budget analysis (1915) develops the mathematical distinction between a change in the relative attractiveness of commodities and the change in purchasing capacity accompanying the price movement. Modern notation expresses this separation with expenditure and compensated demand functions. The original argument and its modern representation belong to the same research lineage, although the later notation should not be projected onto the historical text as a literal transcription.
The observable foundations of consumer analysis became a further research problem. Samuelson’s revealed-choice formulation (1938) examines consistency in choices as observable evidence about consumer behavior. Later nonparametric analysis supplies finite-data tests for rationalization. International household budget research, including Houthakker’s comparison of expenditure patterns (1957), made the systematic relationship between resources and expenditure composition a distinct empirical problem. The historical lines meet in modern demand estimation: a behavioral representation must account for the allocation of a budget, the responses of quantities to changing relative prices, and the ability of observations to discipline those responses.
Point, arc, and finite-change measures
Let denote quantity demanded of commodity , the price of commodity , and the household budget. At a differentiable interior point, the signed Marshallian price elasticity is
The own-price elasticity has ; a cross-price elasticity has . The income or expenditure elasticity associated with the resource variable actually used is
The signed convention is maintained throughout this monograph. Own-price demand is called price-elastic when its negative elasticity has an absolute value above one, unit-elastic when that magnitude equals one, and price-inelastic when it lies below one. A positive own-price derivative calls for a separate explanation. Cross-price and income elasticities retain their signs because these signs distinguish substantively different responses.
Elasticity is invariant to fixed changes of physical and monetary measurement units. If kilograms become grams, the derivative is multiplied by one thousand while the quantity appearing in its denominator is multiplied by the same number. The corresponding cancellation occurs when both price and its derivative are expressed in cents. Unit invariance does not make elasticity invariant to redefining the commodity: gasoline and transportation services, or rice and all grains, are different aggregates.
For two strictly positive observations the midpoint arc elasticity is
The log-change measure is . For a constructed example in which price rises from 10 to 12 and quantity falls from 100 to 80, the midpoint measure equals approximately −1.2222; the log measure is approximately −1.2239. Both summarize a finite interval symmetrically with respect to reversal. The local derivative at either endpoint need not equal either estimate. On a constant-elasticity demand schedule , the log-change measure is exactly for any two positive points on the same schedule. The midpoint measure generally remains an approximation.
A linear demand schedule with has elasticity wherever quantity is positive. Its slope remains constant while the magnitude of elasticity grows with price. Near zero quantity the ratio may become very large, and at zero the ordinary log elasticity is undefined. Researchers should therefore report their functional form and reference point, and use a discrete or extensive-margin measure when the quantity is zero.
Expenditure, revenue, and marginal interpretation
The accounting relation between price and revenue clarifies the practical meaning of a signed own-price elasticity. For expenditure or seller revenue with other relevant arguments fixed,
A negative price elasticity of magnitude below one produces a local rise in revenue when price rises; a magnitude above one produces a local decline. A zero derivative of revenue at unit elasticity is a local statement; global revenue maximization also requires the shape and feasible domain of the demand function. Moving between endpoints by a large price change requires integration of the response along the path, rather than using a derivative at one point as an exact prediction.
This revenue condition does not characterize maximum profit. A firm selling a single product at marginal cost and facing a differentiable downward-sloping demand curve has interior first-order condition . Where the conditions for that optimum hold, . Costs, strategic interactions, multiple products, constraints, and dynamic customer acquisition determine whether that condition is applicable. A change in total revenue alone establishes neither the firm's profit nor household welfare.
Household resources also require an accounting distinction. Total consumption expenditure is an allocation of a budget; disposable money income may be allocated between consumption, saving, and repayment, and a household can use assets or credit. An elasticity estimated with respect to total consumption expenditure is an expenditure elasticity. Recovering the response to current disposable income requires a model or design linking the resource change to the consumption budget. For fixed prices and no other channels, the chain rule expresses the income elasticity as the expenditure elasticity multiplied by the elasticity of total expenditure with respect to income. Income shocks can additionally alter expectations and wealth. The distinction is particularly important for temporary transfers and permanent changes in earning opportunities.
Consumer optimization and dual demand
Consider a consumer maximizing a locally nonsatiated preference representation over nonnegative quantities subject to . Let denote the Marshallian demand. Under conditions ensuring a well-defined optimum and a binding budget, . Its indirect utility is . The dual problem minimizes over the bundles achieving a utility target . Its expenditure function is and its compensated demand is .
At regular points these objects satisfy and . Shephard's lemma gives when expenditure is differentiable. Roy's identity gives where derivatives exist and the marginal utility of resources is nonzero. These relations allow the analyst to connect behavior estimated from budget observations with welfare comparisons expressed through expenditure.
The assumptions warrant explicit attention. The consumer chooses from a specified feasible set, commodities have positive prices, preferences are held fixed for the comparative static, and money income is the resource whose variation is controlled. A rationed consumer may face a binding quantity ceiling; a household with an irregular tariff may face a kinked budget; a consumer lacking access to a product cannot necessarily purchase it at the recorded market price. Duality can be extended to other environments, but the usual formulas describe the stated linear-budget model. A scientific estimate records any departure from that environment.
The Slutsky decomposition
Differentiate with respect to , holding the utility target fixed. At the common optimum, and hence
Define the compensated elasticity and the expenditure share . Multiplication by yields the central elasticity identity:
The compensated response holds the preference level fixed through a change in expenditure. The Marshallian response holds nominal resources fixed. The difference equals the expenditure share of the commodity whose price changes multiplied by the resource elasticity of the commodity whose demand is measured. For a household spending a large fraction of its budget on a commodity, the purchasing-power channel may have quantitatively important effects throughout the budget.
For a normal good, , the own-price income effect reinforces compensated own-price contraction. For an inferior good, , that effect opposes the contraction. The identity thus locates Giffen demand inside ordinary consumer theory when the positive purchasing-power effect exceeds the absolute compensated effect. The argument is local, with preferences and other prices fixed; it applies at the common reference point for the two derivatives. Compensating a household by its original basket cost after a finite price change is a different finite experiment from holding utility exactly constant, although the compensation rules coincide to first order.
Budget restrictions and integrability
Differentiating with respect to expenditure and prices establishes restrictions for a complete demand system:
Scaling all prices and expenditure proportionately preserves the linear budget set. Homogeneity of degree zero gives
The expenditure function is concave in prices because it is the pointwise infimum of linear expenditure functions of prices over the feasible utility set. Its twice-differentiable Hessian is symmetric and negative semidefinite. Through Shephard's lemma the compensated substitution matrix inherits these restrictions, so compensated own-price elasticities are nonpositive. Symmetry in elasticity notation is weighted:
It is the derivative matrix, rather than the unweighted elasticity matrix, that is symmetric. Negative semidefiniteness requires for every vector , a stronger condition than a nonpositive diagonal. Compensated homogeneity and the adding-up identity give and . In applied work these restrictions concern a complete allocation, using consistent quantities, positive shares, and the appropriate reference point. Imposing them on an arbitrary subset without modeling the unobserved complement changes the research object.
A constructed two-good Cobb–Douglas case makes these restrictions transparent. Let and . Both expenditure elasticities are one. The Marshallian matrix has diagonal entries −1 and off-diagonal entries zero. Slutsky's identity gives compensated rows and . The unequal off-diagonal elasticities reflect different expenditure shares; their share-weighted values both equal 0.24. The compensated rows sum to zero and the weighted columns sum to zero. This example also illustrates that a gross cross-price elasticity of zero can coexist with a positive compensated cross-price response.
Substitution, complementarity, and Giffen demand
For a price rise in commodity , a positive gross cross-price elasticity of commodity means that demand for rises at fixed nominal expenditure; a negative one means that it falls. Compensated cross-price effects answer the different question with utility held fixed. A consumer compelled to spend more on an essential commodity may reduce several other purchases even when substitution toward those purchases would occur under compensation. The expenditure share determines the scale of this budget channel relative to the resource response of .
In a two-good complete system, compensated homogeneity and negativity imply nonnegative off-diagonal compensated derivatives. With three or more goods, compensated complements can exist because a relative price movement can make an entire combination less attractive. Complementarity and substitutability can therefore depend on whether preferences, technology, or a particular demand derivative supplies the definition. A fixed-ratio consumption technology permits still another interpretation, involving limited substitution within a joint use.
A constructed Giffen case illustrates the threshold without claiming the existence of such coefficients in a particular population. Suppose a two-good setting at a regular interior point has shares , resource elasticities and compensated rows and . These local coefficients satisfy the weighted adding-up, homogeneity, and compensated symmetry restrictions; negative semidefiniteness follows for the two-good substitution matrix with equal expenditure shares. Slutsky's equation gives the first good an own-price elasticity of . The second good has own-price elasticity −1.6. The example demonstrates local compatibility of positive own-price demand with the restrictions; it does not supply a global utility representation or an empirical estimate.
Many inferior goods have negative own-price demand because the resource response is too small to outweigh substitution. A Giffen response is also distinct from a demand increase caused by a price signal of quality, prestige, or expected future scarcity, since those mechanisms change information or other arguments of demand. In data, a positively sloped equilibrium association can arise from a demand shock without any positive own-price consumer derivative. The empirical question concerns the source of price variation and the purchasing capacity and opportunities actually affected.
Engel curves and the allocation of additional resources
An Engel curve describes demand for a commodity as resources change at fixed prices and other relevant conditions. Let be expenditure on commodity and . Differentiating the share gives
A declining expenditure share is thus compatible with increasing physical consumption when . For example, a household whose food expenditure rises at fixed prices from 30 to 36 units while its total budget rises from 100 to 150 increases food purchases by twenty percent, yet its food share falls from thirty to twenty-four percent. The log-interval elasticity is approximately 0.4497. Calling food inferior from the share decline would confuse the level with the allocation fraction.
The marginal budget share is . It measures the fraction of an additional unit of budget allocated to the good locally. The average share and marginal share coincide under unit expenditure elasticity. They can differ considerably at subsistence levels and for goods that become affordable at higher incomes. For a complete budget the marginal shares sum to one. Some can be negative in an inferior-good region, requiring others collectively to exceed one.
Cross-sectional Engel curves vary with household composition, household size, location, and relative prices. A curve fitted across households at one survey date need not describe the trajectory of one household as its lifetime resources change. Cohort differences, household formation, and transitory income shocks can contribute to the cross-sectional gradient. Equivalence scales offer a means of comparing households with different needs under specified assumptions; a scale itself constitutes an empirical and normative choice rather than a directly observed invariant demand response.
Subsistence and nonhomothetic preferences
A useful demonstration of nonhomothetic demand is the linear expenditure system derived from a Stone–Geary specification. For feasible resources above subsistence expenditure, let be a minimum consumption parameter and with . The interior Marshallian allocation is
The marginal budget allocation is constant at , while the expenditure share varies with the excess of budget over the subsistence bundle. The expenditure elasticity equals . Goods whose initial shares exceed their marginal allocation consequently have elasticity below one; other goods can have elasticities above one. As discretionary expenditure grows without bound in this specification, the shares approach the marginal budget shares. The model illustrates the mechanism through which requirements and discretionary consumption generate systematic income variation.
This specification with nonnegative minima and positive marginal shares produces positive income derivatives. It cannot reproduce inferior demand or the Giffen mechanism in its regular feasible region; those outcomes require different preference or constraint structures. Its fixed marginal shares can also become too restrictive over a wide income range. The usefulness of the example lies in the transparent separation between committed quantities and discretionary allocation, not in its status as a universal empirical law.
A homothetic representation instead has Engel curves proportional to expenditure at a fixed price vector, whenever interior demands are unique. The expenditure elasticity of every positive quantity in such a model is one. A given elasticity of substitution, including in constant-elasticity-of-substitution preferences, concerns a different derivative: how a ratio of quantities changes in response to a relative marginal valuation under particular conditions. It does not imply a universal own-price elasticity. Budget shares enter the conversion between substitution properties and commodity demand derivatives. This difference is important when preference parameters estimated from a structural model are reported alongside policy-relevant elasticities.
Flexible demand systems
A demand system estimates a set of interdependent allocations instead of treating each commodity as an isolated equation. The Almost Ideal Demand System of Deaton and Muellbauer (1980) expresses expenditure shares as functions of log prices and log real expenditure. In the exact nonlinear formulation,
where
Adding up requires , and for every . Homogeneity requires for every . Symmetry requires . Given the symmetric price-index representation, . By differentiating at fixed prices or fixed expenditure, the elasticity equations follow:
The Kronecker indicator equals one when and zero otherwise. Hicksian effects are obtained by the Slutsky identity. Using observed expenditure shares in place of the exact derivative of the price index is a different approximation. Estimation and elasticities computed under a linearized price index must therefore document their conventions separately. The original application to British data demonstrates empirical testing of theoretical restrictions; flexibility in representation does not guarantee that the restrictions will be accepted by observed data.
The quadratic Engel-curve extension of Banks, Blundell, and Lewbel (1997) adds nonlinear expenditure dependence to the share equations. With suitable functions and , and , a QUAIDS share can be represented as
where and . At fixed prices, the expenditure elasticity becomes . Its curvature permits the resource response to vary across the expenditure distribution more freely. That additional flexibility matters when extrapolating an average estimate to the poor or rich. Positivity of shares, monotonicity, and compensated curvature must be checked at economically relevant points; algebraic adding-up and symmetry alone do not ensure global regularity. Data scarcity and overfitting may also make the nonlinear terms poorly estimated.
Revealed preference and the information in finite samples
Observed choices can test restrictions without committing to a fully specified smooth demand system. For observations satisfying a linear budget with the chosen bundle exhausting expenditure, an Afriat-type test seeks numbers and such that, for all pairs of observations,
Under the standard finite-observation conditions, feasibility supplies a concave, monotone utility rationalization; violations of the generalized axiom of revealed preference reject the exact consistency requirements. Varian’s nonparametric treatment (1982) develops this research program, building on Afriat's theorem. Rationalization is a significant empirical restriction, yet a finite set of consistent choices generally admits many representations with different local derivatives between observations.
A constructed example clarifies the informational boundary. If two household observations have identical relative prices but different expenditures, the data can reveal one finite Engel-curve segment for each observed quantity. They contain no direct independent variation of relative prices from which to learn a cross-price derivative. A smooth utility representation may produce one implied cross-price response, but competing rationalizations can differ. A fitted structural elasticity adds information from assumed curvature and functional form that the finite data alone do not identify.
Observed inconsistencies can reflect changing tastes, errors in prices or quantities, liquidity constraints, household composition changes, or failures of the chosen budget model. Empirical implementations may allow explicitly defined measurement-error tolerances or behavioral departures. A researcher should report the exact test before and after tolerance, the source of tolerance, and the effects on the set of feasible responses. Convenience in obtaining a single coefficient is insufficient justification to hide the range of admissible demand functions.
Heterogeneity and population aggregation
Aggregate demand is the sum of the quantities of the decision makers in the specified population. If all households face the same changing price, total resources and composition are held fixed at their household-specific baseline conditions, and each demand is differentiable, the aggregate elasticity equals the quantity-weighted average of household elasticities:
The weights are shares of the quantity of commodity , not population shares or total household budgets. For two equally numerous groups consuming 20 and 80 units respectively, with own-price elasticities −2 and −0.25, the aggregate elasticity is −0.6. The unweighted average, −1.125, describes a different statistic. Reporting the latter as total market responsiveness would misstate the quantity impact.
Income aggregation requires the distribution of the resource change. If every household's budget increases proportionately, the same quantity weighting can be used for resource elasticities. An equal transfer in monetary units produces proportionally larger budget changes for poorer households; its aggregate response depends on each household's baseline budget and marginal demand. A transfer targeted to a subgroup changes the weights again. Even when prices remain constant, the policy elasticity of total demand with respect to aggregate income is conditional on the allocation of the increment among households.
Population composition is another state variable. Migration, demographic change, access to retail outlets, and participation may change the number of potential purchasers. For a fixed market boundary, differentiating mean demand and the number of consumers distinguishes population growth from behavioral response. A representative consumer can supply convenient mathematical closure, but its demand properties should be derived under explicit aggregation assumptions or treated as an approximation. Individual regularity alone does not establish a stable representative-income response when distribution changes.
Conditional demand and product aggregation
The construction of commodity groups alters measured responsiveness. A narrow brand competes with close alternatives, whereas a broad food category includes internal substitutions that leave total group expenditure or quantity relatively stable. Several products with different physical units cannot always be summed as a meaningful homogeneous quantity. A quantity or volume index supplies one aggregate, and a particular price index supplies its companion. Their weights and quality treatment must be documented.
A two-stage budget model illustrates the distinction between conditional demand and demand in the complete budget. With defensible weak separability, a household can allocate expenditure among groups and subsequently allocate a group budget among commodities. An elasticity within the food budget holds that food allocation fixed. The total elasticity also incorporates the response of the food allocation to the price disturbance. By the chain rule, the change in the log of within-group quantity equals its direct conditional price response plus its conditional expenditure elasticity times the change in log group expenditure, provided the remaining conditioning arguments are handled consistently.
Unconditional coefficients are therefore local derivatives of the completed allocation problem. Researchers who compare a conditional estimate for one country with an unconditional estimate for another may attribute a difference to preferences even when it comes from the budget boundary. USDA’s international demand documentation distinguishes these stages in historical global food-demand estimates. Its categories and price conventions provide a valuable example of the need to match the estimand before international comparison.
Participation, product choice, and quality adjustment
For goods purchased intermittently or available only to a subgroup, write mean purchases as . Under differentiability and a positive mean, the elasticity of mean purchases is the sum of the elasticity of the purchase probability and the elasticity of conditional mean purchases. The first is an extensive-margin response; the second describes a changing conditional group as well as individual quantities. If a price movement causes low-quantity consumers to leave the market, the mean quantity among remaining consumers may increase even when each continuing consumer purchases less.
Discrete product selection also requires attention to the distinction between own-product substitution and exit to an outside option. A consumer choosing one durable good from many may respond to a price movement by selecting another brand, delaying acquisition, or choosing no product. Berry, Levinsohn, and Pakes’s automobile study (1995) estimates differentiated-product demand and equilibrium supply using product-level data and instruments. The structure permits substitutions to depend on product attributes and heterogeneous consumers. The resulting elasticities remain conditional on the modeled choice set, market definition, demand specification, and instrument validity.
Households can change product quality, package size, store, and delivery arrangements in response to prices and income. The expenditure-to-quantity ratio for a group of commodities is a unit value that includes quality composition. A household with rising resources may buy a higher-quality version of the same commodity; group expenditure can rise even if physical quantity stays constant. Conversely, substituting to low-quality varieties can reduce measured unit value in a period of inflation. The differentiation between posted price and unit value is essential to interpreting both price and resource elasticities.
Adjustment horizons and changing opportunities
The horizon of response changes the feasible adjustment channels. A consumer may rearrange current purchases quickly but require time to change appliances, vehicles, housing, or work locations. Stockpiling may raise purchases today and reduce them subsequently. Durable goods provide services over time; the purchase flow and the service flow exhibit different elasticities. For an anticipated tax change, purchases can shift before implementation, making a simple before-and-after comparison misleading.
In a stylized stable dynamic log-demand equation with lagged quantity,
the impact price elasticity is and the steady-state derivative for a permanent common log-price change is , assuming and the model's other inputs held fixed. This derivation is conditional on a correctly specified stable adjustment process. Transitory price changes, nonstationary budget paths, expectations, and shifts in the capital stock can require different models.
Havranek, Irsova, and Janda’s gasoline synthesis (2012) reports publication-selection-adjusted mean estimates of −0.09 for the short run and −0.31 for the long run. These are historical meta-analytic results as reported in the article abstract, not newly estimated coefficients or a guarantee for any present-day jurisdiction. Their distinction illustrates the significance of horizon and study selection. A defensible policy application should match the fuel, period, taxation structure, vehicle stock, and behavioral margins of the target population.
Gasoline demand and the resolution of observations
Hughes, Knittel, and Sperling’s gasoline estimates (2008) compare two U.S. high-price periods. Their publisher abstract reports short-run price elasticities ranging from −0.034 to −0.077 in 2001–2006 and from −0.21 to −0.34 in 1975–1980. These ranges describe the models and periods analyzed. The comparison establishes a difference in estimated responsiveness without isolating a single causal explanation for the difference. Changes in vehicle stocks, commuting patterns, incomes, and substitution possibilities are potential mechanisms that require distinct evidence.
Levin, Lewis, and Wolak’s high-frequency analysis (2017) employs prices and expenditure data at daily frequency for 243 U.S. metropolitan areas over 2006–2009. The authors find that distinguishing purchase timing from gasoline demand and changing spatial and temporal aggregation can yield substantially different elasticity estimates. This research complements the evidence on adjustment horizons: frequency of observation and economic length of response are related but distinct. A daily purchase can reflect both current consumption and the timing of a refill; a monthly aggregate combines prices encountered by heterogeneous consumers at different times.
The three gasoline studies discussed here illustrate different dimensions of empirical variation. The historical period changes the economic setting; the adjustment horizon changes possibilities for capital and behavioral adaptation; and data aggregation changes what the researcher observes. These dimensions should be assessed jointly before transporting a fuel elasticity into a tax simulation. A model of marginal tax incidence over a fixed interval calls for a corresponding market quantity response, whereas a transport transition extending over years requires a representation of vehicle and infrastructure adjustment.
Price information and economic constraints
Prices act through budget opportunities and through the conditions under which consumers recognize those opportunities. Consumers may observe a shelf price while overlooking a delivery charge, a periodic subscription renewal, or a future rebate. A posted price is then an incomplete description of the decision environment. The elasticity with respect to the posted component of a nonlinear tariff may differ from the effect of changing its marginal per-unit rate or its total expected cost.
An observed response can also reflect binding availability constraints. If a household consumes exactly the amount allocated by a rationing rule, a change in the monetary price within a range can leave quantity unchanged even when its unconstrained demand would respond. Informal market prices, queues, travel costs, and quality differences may determine the effective price. When the official price and these other costs move in different directions, an estimate indexed solely to the official price can have limited interpretability.
Price elasticity estimates should state whether the change is salient to the consumer and whether its consequences are immediate or expected. For demand under uncertainty, income changes can alter liquidity and expected future resources differently, especially when access to borrowing or savings is restricted. A temporary discount and a permanent cost reduction are distinct interventions. The empirical design must isolate their effects before a common behavioral parameter is asserted.
Medical demand and nonlinear prices
Medical demand often occurs under a contract with deductibles, cost sharing, coverage limits, and a provider network. The current payment for one service and the expected marginal price of future services can depend on cumulative spending. A patient near an out-of-pocket maximum can face a different incentive from another patient with the same contract who has incurred no expenditure. Choices may additionally involve uncertainty over illness, the advice of medical professionals, and access constraints.
Einav and Finkelstein’s insurance review (2018) examines credible empirical designs and the contribution of structural modeling to counterfactual evaluation. Their comparison of health insurance experiments, including RAND and the Oregon Medicaid lottery, helps distinguish variations in cost sharing from changes in insurance access. Evidence of a behavioral response to coverage cannot automatically be interpreted as a conventional own-price elasticity for an individual service. The intervention may change several economically relevant conditions at once.
A policy analyst must distinguish utilization from health outcomes, spending from resource costs, and the immediate behavioral response from the value of insurance against uncertainty. A measured increase in treatment following a change in financial incentives does not settle the distribution or health benefits of the change. The institutional price schedule and the nature of the intervention determine how far a static demand derivative can inform a welfare analysis.
Causal identification of demand
The price observed in a transaction is often jointly determined by demand and supply. Consider a local log-linear demand equation with . If demand shocks raise both equilibrium price and equilibrium quantity, then . An ordinary least squares regression estimates the structural coefficient plus the ratio of this covariance to the variance of log price, under the simple regression assumptions. The resulting positive bias can make a strongly negative underlying demand response appear weak or positive. Alternative supply responses and measurement errors can change the direction of bias.
The discussion of counterfactuals in Berry and Haile’s demand estimation foundations (2021, revised 2022) emphasizes the central role of reliable quantitative demand estimates in studying market power, taxation, and welfare. Instrumental-variable estimation uses a source of variation correlated with price and excluded from the demand disturbance, conditional on the maintained design. Cost shocks can shift supply, but a cost shock that also changes product quality, availability, or household income may fail the exclusion restriction. First-stage strength measures relevance rather than validity. With heterogeneous responses, the IV estimand can weight particular changes and groups; a single scalar instrument need not identify the entire vector of cross-price derivatives.
Panel methods remove time-invariant unobserved differences under appropriate specifications. They do not automatically remove time-varying shocks that drive both prices and quantities. Difference-in-differences designs need an untreated counterfactual evolution, and event-study estimates require attention to anticipating behavior, staggered treatment, and changing composition. Spatial comparisons can exploit local price variation, but migration, common infrastructure, store access, and correlated local preferences can undermine the intended comparison. Each method earns a causal interpretation from its specific assumptions and supporting evidence.
Randomized price offers can supply a clearer price perturbation when assignment is independent of potential outcomes. The assignment effect measures the intervention as delivered; if participants do not take up the offer, the effect of actual price paid requires additional assumptions or identification through assignment. Spillovers between treated and untreated households, resale, rationing, and supply responses can alter the counterfactual. Price variation and budget variation are both required for identifying a complete demand system unless credible structural restrictions supply the missing information.
Measurement of household resources and prices
Household expenditure data are useful because they describe multiple allocation decisions under resource constraints. The U.S. Consumer Expenditure Surveys guide describes complementary Interview and Diary components and the limits of public-use information. The Eurostat household budget microdata documentation establishes the availability of research-access survey waves, but differences in survey implementation and the timing of price measures require harmonization. Neither documentation source alone yields a causal demand elasticity.
A household survey can combine recalled and recorded expenditure, imputed own production, gifts, in-kind transfers, and durable acquisitions. Survey design experiments summarized in the World Bank’s consumption measurement program examine recall and diary instruments, nonstandard quantities, and food consumed away from home. When prices are inferred from survey amounts rather than independently recorded, errors in expenditure and quantity can enter both the unit-value regressor and the dependent variable. The resulting correlation may be mechanical as well as behavioral.
Researchers should distinguish disposable money income, comprehensive household resources, consumption expenditure, and the budget allocated to a particular group; choose an observation period that connects expenditures to relevant prices; convert quantity units carefully; and record whether price observations are transaction prices, posted prices, spatial indices, or unit values. Households with no observed purchase need a participation model or an explicit statement of the population restricted to buyers. Dropping all zeros and estimating a log-linear equation can change the research population and obscure response at the extensive margin.
Unit values and spatial variation in Africa
Deaton’s study of Côte d’Ivoire (1988) uses 1979 household data to examine how spatial variation and within-cluster information can help distinguish price effects from quality choice and measurement error. The argument depends on assumptions about common local prices, observable and unobservable household differences, and the way quality changes with resources. It gives a disciplined alternative to interpreting each household's expenditure divided by quantity as an exogenous market price.
A later investigation by Nsabimana using Tanzanian household data (2024) compares community price measurements with unit-value approaches in a flexible food-demand system. Its reported resource elasticities are comparatively stable across these measurement approaches, whereas estimated price responses are sensitive to the chosen proxy. This result concerns the study's data and model; it identifies a practical source of specification uncertainty rather than establishing a universal ranking of price proxies. Community prices can themselves differ from transactions in timing, location, and product quality.
The Ethiopian food-demand study of Tafere and coauthors (2010) uses 2004/05 household income, consumption, and expenditure survey data. Its official abstract establishes a country-specific demand-system application. The study is included here as evidence of the geographic and commodity-specific character of demand estimation; its numerical elasticities and full estimation diagnostics have not been independently examined. Together these cases motivate a requirement for paired sensitivity analysis using alternative defensible price and resource definitions whenever the data support it.
Experimental evidence of Giffen behavior in China
Jensen and Miller’s subsistence experiment (2008) investigates staple-food demand using randomized subsidies in poor Chinese households. The study reports strong evidence of Giffen behavior for rice in Hunan and more limited evidence for wheat in Gansu. Its interpretation concerns subsistence constraints and heterogeneity among poor households rather than the average behavior of all consumers of these commodities. The subsidy changes effective purchase opportunities and enables comparison across assigned treatment conditions.
The theoretical mechanism can be understood as follows. A household consuming a staple to satisfy basic needs also seeks consumption of other desirable foods. If its resources are sufficiently restricted, a higher staple price can reduce the feasible quantity of the more desirable foods, forcing an increase in reliance on the staple. For a household at a different level of deprivation or with different alternative foods, substitution may dominate or the feasible adjustment may differ. These conditional possibilities explain why a positive own-price elasticity can appear in one population segment and disappear upon aggregation.
The experiment supplies evidence about the stated price intervention in the studied populations. Extrapolating to urban households with more diverse food access, a different staple, an unrestricted subsidy, or a prolonged economy-wide price increase requires a new assessment of constraints and external validity. The evidence supports the existence of the mechanism under relevant circumstances; it does not support treating staple demand as universally upward-sloping. This distinction is crucial for food security policy, where broad average estimates may conceal the response of the most vulnerable groups.
International expenditure patterns and purchasing power
Houthakker’s international evidence (1957) compares household budget patterns across approximately thirty countries and about forty surveys. The comparative tradition shows why the composition of expenditure can be related systematically to resource levels while remaining conditioned by differences in relative prices, customs, demographics, and survey institutions. Engel's law concerns the regular empirical tendency of the food share to decline as resources increase, rather than an axiom about every household under every price system.
Modern cross-country demand research extends the comparative approach with internationally assembled price and expenditure information. The USDA study using 2005 ICP data (Muhammad et al. 2011) covers 144 countries. Its historical estimates permit systematic comparisons of broad expenditure categories within the study's conventions. They concern a specific reference year and estimation framework; the associated data documentation indicates that the series is no longer updated. Its coefficients cannot be read as estimates for the current world economy merely because the dataset remains accessible.
Zereyesus and coauthors' (2025) updated international elasticity estimates use 2017 ICP observations for 176 economies and an AIDS framework. Their data section distinguishes economies from countries and reports unweighted country-group means of food shares. The change in sample and specification relative to earlier work means that comparisons are not controlled longitudinal estimates of preference changes. Even a harmonized commodity code can conceal evolving qualities and different consumption environments.
The global food and non-food spending assessment of Nzayiramya, Muhammad, and Baffoe-Bonnie (2025a) analyzes the 2021 ICP round for 173 countries with QUAIDS. A separate publisher correction notice (2025b) dated October 9 corrects a word in the original article rather than its data coverage or model. The studies' distinct vintages and specifications offer a research program for future, harmonized comparisons. This monograph uses the research scope and estimation frameworks rather than claiming to have replicated their reported elasticities.
The International Comparison Program’s 2021 release (World Bank 2024) provides purchasing-power parity and price-level comparisons for the stated reference year. These are designed for international comparison of economic aggregates. A national purchasing-power parity is neither an individual household budget nor an independently assigned marginal commodity price. Cross-country demand studies require compatible product classifications, purchasing-power adjustments, demographic definitions, and price variation. Even with these harmonizations, observed differences between countries may reflect institutions or preferences as well as expenditure effects.
International comparison therefore benefits from reporting responses at comparable locations in resource distributions and distinguishing within-country variation from between-country variation. A country-average coefficient estimated from national accounts can differ from the average household-level elasticity, both because of aggregation weights and because the design identifies a different relationship. The value of comparative research increases when the differences in estimands are explicit and the practical consequences of that mismatch are examined rather than averaged away.
Household demand across the European Union
Gehrke and Jin’s EU food-demand report (2026) examines representative cross-sectional household survey data for the 27 member states. The institutional abstract describes an EASI demand system with weak separability between consumption and other expenditures, and reports heterogeneity by household income and member state. At the aggregated EU level the study reports normal food categories and negative own-price elasticities; poorer households are more responsive to prices overall in its estimates. The research is a report, and this review relies on its institutional abstract rather than independently estimating its coefficients.
These data complement international national-account comparisons because they describe heterogeneity within the household population. Their cross-sectional character still limits causal interpretation of resource gradients and price variation without a supporting design. The separability and resource boundaries deserve particular attention when comparing the estimates with complete-budget systems. A regional policy analysis can gain from representing variation by household income while using consistent definitions of food groups, exposure, and total expenditure.
Retail organization and demand in Mexico
The entry of new retail formats can alter prices, product assortment, travel opportunities, and existing retailers' behavior. Atkin, Faber, and Gonzalez-Navarro’s Mexico research (2016 manuscript; final article 2018) examines the welfare consequences of retail globalization through changes in household shopping opportunities and costs. A reduction in the cost of living can arise from direct access to a new store and responses elsewhere in the market. Heterogeneous households can benefit through different channels.
This case illustrates a limitation of elasticity calculations performed with a fixed commodity and seller set. A new store can create products or shopping arrangements that were previously infeasible; its entry can change reservation prices, variety, transportation costs, and the price vector faced by other consumers. An own-price elasticity estimated for an incumbent product before entry does not by itself evaluate welfare changes associated with the changed opportunity set.
An analyst can distinguish among the price effect on the continued purchase of a particular product, substitution across stores, participation in the newly available retail format, and the general equilibrium response of local sellers. Each component can inform a welfare analysis when its definition and estimation design correspond to the policy question. This is a different task from assigning all consequences to one demand elasticity on a fixed schedule.
Estimation, uncertainty, and validation
Demand functions can be estimated through constant-elasticity log models, share systems, discrete-choice structures, semiparametric representations, or experimental contrasts. Models with different local fit may give very different predictions far from the observed price support. A small residual sum of squares does not guarantee credible derivatives; derivatives can be sensitive to smoothing, weak instruments, and the curvature of a fitted function. Nonparametric and flexible forms reduce some restrictions at the cost of increased data requirements.
Uncertainty in demand elasticity often involves ratios and nonlinear transformations of estimated parameters. If an elasticity is a differentiable function , the delta method approximates its variance by . The covariance among estimated parameters matters, especially in a complete demand system. Bootstrapping can propagate sampling and estimation variation through the calculation when resampling respects the design. Clustered price variation or randomization calls for uncertainty calculations at the appropriate level; a large number of household observations does not imply many independent price shocks.
A demand-system validation should examine share predictions, adding-up identities, homogeneity, weighted symmetry, and the eigenvalues of compensated substitution matrices over the economically relevant domain. It should compare restricted and unrestricted estimates, test sensitivity to expenditure and price measurement, examine outliers and extreme fitted shares, and evaluate predictions under holdout observations or credible intervention data. Local regularity should be tested near the policy counterfactual rather than inferred solely from global functional-form labels. A design with a small number of policy shocks or a narrow range of observed prices may justify broad uncertainty intervals or partial identification.
A transparent empirical report preserves the distinction between sampling uncertainty and specification uncertainty. An interval around a coefficient conditional on one model cannot automatically account for an invalid instrument, missing quality dimension, or alternative resource definition. Sensitivity under materially different credible specifications is an additional result. When plausible estimates disagree in sign, the scientific conclusion may concern the inadequacy of identifying information rather than an average of the opposing signs.
Tax incidence, pricing, and public policy
Demand elasticity is one part of a tax-incidence calculation. Consider a competitive market near an initial no-tax equilibrium at common price , with specific tax creating . Let and denote the respective local demand and supply elasticities. Differentiating market clearing gives the local fraction of a marginal tax passed to consumers:
In the constructed case and , this fraction equals 0.75. Consumer prices rise by three quarters of a marginal monetary tax unit, seller receipts per unit fall by one quarter, and quantity contracts. The result applies at the stated common-price baseline for competitive supply and demand with the maintained assumptions. At an existing tax wedge the slopes must be evaluated at their corresponding prices. Market power, nonuniform taxes, product quality changes, and salience can change the result.
The tax revenue derivative combines the quantity tax base and the response of that base as the tax changes. Its sign is not determined solely by the demand elasticity; supply matters through equilibrium, and taxes on substitutes can change the outcome. Distributional analysis further asks who consumes, who supplies, who pays the tax, and who receives any revenue-financed transfers. Elasticity is a response to an economic change, not a complete measure of fiscal welfare or distributive justice.
A price subsidy may improve access to essential goods while expanding consumption by a relatively small amount in the aggregate. It may also produce heterogeneous quantity effects, including different responses for households near subsistence. A cash transfer affects all demands through the budget, while a commodity-specific subsidy changes relative prices as well as resources. These instruments have different behavioral and fiscal consequences even when their immediate cost is the same. Policy comparison therefore requires both the complete response system and an explicit welfare criterion.
Welfare and the economic cost of living
For a change from price vector to with nominal expenditure held at , let and . Define compensating variation as the positive compensation required to restore initial utility after an adverse price change, . Define equivalent variation as the positive amount that could be removed at initial prices to reduce utility to its post-change level, . These sign conventions give positive losses for an adverse change and negative losses for an improvement.
Shephard's lemma implies that the derivative of expenditure along a price path holding utility fixed is the compensated quantity vector. Exact cost-of-living comparisons therefore involve the integral of compensated demand, or evaluation of the expenditure function. A local cost approximation is useful for small changes. Large changes introduce substitution and curvature, so multiplying the initial basket by the whole price change can differ from exact compensation.
For a constructed two-good Cobb–Douglas consumer with baseline expenditure 100 and expenditure share 0.4 on the good whose price increases by twenty-five percent, holding the second price fixed, the expenditure index necessary to maintain baseline utility increases by a factor . Consequently and . The initial basket would cost an additional 10 without substitution. A demand decline or an observed expenditure increase alone would not recover either exact welfare measure. This example is entirely constructed and illustrates the distinction among quantity response, actual spending, and welfare cost.
Household welfare comparisons across heterogeneous populations require further decisions concerning interpersonal aggregation, distributional weights, and the use of public revenue. A money-metric measure for one household is conditional on its preference assumptions and reference prices. Consumer surplus provides useful approximations under particular conditions; compensated or equivalent variation gives the appropriate exact dual comparison when the expenditure function is available. An elasticity estimate informs welfare analysis through the underlying demand system; it cannot substitute for the whole welfare calculation.
Integrated analytical synthesis
The preceding analysis identifies four distinct economic objects that often share the same name in empirical reports. The compensated individual response describes substitution at a fixed preference level. The Marshallian household response describes the reallocation of a specified nominal budget. The population response aggregates household quantities under a specified distribution of prices, income changes, and opportunities. The equilibrium response incorporates supply, competition, market entry, and other adjustments. Their relations are derived from consumer duality, budget aggregation, and market clearing; their separation is a synthesis of established economic analysis rather than an additional theory.
An important consequence follows for evidence interpretation. Observed price changes and quantities principally characterize the experiment actually realized by consumers and sellers. If relative prices alone change under a controlled intervention, the resulting response may identify a Marshallian derivative. If the intervention compensates each household for changes in its preference level, it can approach a compensated derivative. If the change arises from a market-wide tax, a structural demand response needs to be separated from supply and institutional responses. If cross-sectional price differences include quality choice, the measurement itself changes with the allocation. The appropriate econometric design is therefore determined by the precise counterfactual, not by the convenience of regressing one variable on another.
A second consequence concerns variation across resource levels. The same compensated own-price response can coexist with very different Marshallian responses because expenditure shares and resource elasticities vary. A household near subsistence may devote a large share to basic necessities, so changes in their prices can produce substantial budget effects throughout consumption. A more affluent household can have a smaller budget exposure even when its physical response to relative prices is similar. Studies comparing countries or regions should consequently distinguish changes in preference-related substitution from changes in exposure through budget shares and resource distribution.
A third consequence concerns the transfer of estimates. An estimate can be transported only to the extent that its commodity boundary, resource definition, price intervention, adjustment horizon, heterogeneity, and relevant institutional conditions remain comparable or can be adjusted through a credible structural representation. A literature mean characterizes the studies and weighting from which it was computed. A local experimental coefficient identifies the behavior affected by its intervention. A flexible structural estimate can predict a new policy if its maintained relationships remain valid over the new domain. The scientific task is to establish those relations explicitly and to preserve meaningful uncertainty where the evidence does not determine them.
The practical reporting implications are concrete. An elasticity should be accompanied by the population and sampling frame; baseline prices, quantities, and relevant budget shares; the exact resource measure; the form and compensation of the price perturbation; conditional or complete-budget status; the adjustment horizon; the source of identifying variation; the treatment of quality, zero purchases, and composition; the uncertainty calculation; and the limits of extrapolation. For a demand system, the restrictions and regularity of the complete matrix should also be reported. These elements make the magnitude and sign economically interpretable and make comparisons among studies scientifically meaningful.
Research frontiers and empirical priorities
Richer household and transaction records can improve measurements of price exposure, purchase frequency, product switching, and household heterogeneity. An incomplete observation of transactions, however, can exclude cash purchases, informal markets, consumption outside a digital platform, and purchases made by other household members. A research design needs to distinguish improved precision within the observed channel from a complete representation of the consumption budget.
A promising empirical comparison would estimate related demand systems under a common commodity classification using several ICP vintages and compatible household survey samples, explicitly separating data revisions, price-index definitions, sample composition, and nonlinear expenditure dependence. It would also assess transportability in the neighborhood of large price or income changes, where the usual local approximation may fail. Credible experiments and independently measured prices remain especially valuable for identifying the conditional behavioral relationships required by these models. Such a comparison is a proposed research direction; this monograph does not claim to have estimated it.
Conclusion
Price and income elasticities express measurable relationships within economic choice. Their mathematical structure connects relative price responses, resource responses, and expenditure shares. The Slutsky identity establishes the relationship between compensated and Marshallian elasticity; adding up, homogeneity, symmetry, and compensated negativity discipline complete demand systems. Engel curves and heterogeneous budget exposure explain why a response can vary across the resource distribution, while aggregation, extensive margins, quality choice, and adjustment horizons clarify why different demand measures need not share a single coefficient.
Empirical identification follows the economics of the response. Market price movements can be endogenous; survey unit values can embody quality; the resource variable can describe expenditure rather than income; and reforms can change the available choice set. Randomized interventions, credible instruments, carefully constructed household data, and tested structural models each illuminate a defined aspect of demand behavior. Comparative evidence is most useful when its study geography, time, consumption category, and estimation assumptions remain explicit.
The original synthesis of the monograph establishes that the scientific value of an elasticity lies in the combination of its numerical response and its well-defined counterfactual. Precise identification of the response object connects the mathematics of demand to empirical observations and public decisions. This foundation supports research on consumption, pricing, taxation, distributional welfare, and market organization while respecting the limits of each evidential design.
References
Atkin, David, Benjamin Faber, and Marco Gonzalez-Navarro. 2016. “Retail Globalization and Household Welfare: Evidence from Mexico.” Author manuscript, May. Final publication: Journal of Political Economy 126 (1), 2018: 1–73. DOI: 10.1086/695476.
Banks, James, Richard Blundell, and Arthur Lewbel. 1997. “Quadratic Engel Curves and Consumer Demand.” Review of Economics and Statistics 79 (4): 527–539. DOI: 10.1162/003465397557015.
Berry, Steven T., and Philip A. Haile. 2021, revised 2022. “Foundations of Demand Estimation.” Cowles Foundation Discussion Paper 2301.
Berry, Steven, James Levinsohn, and Ariel Pakes. 1995. “Automobile Prices in Market Equilibrium.” Econometrica 63 (4): 841–890.
Deaton, Angus, and John Muellbauer. 1980. “An Almost Ideal Demand System.” American Economic Review 70 (3): 312–326.
Deaton, Angus. 1988. “Quality, Quantity, and Spatial Variation of Price.” American Economic Review 78 (3): 418–430.
Einav, Liran, and Amy Finkelstein. 2018. “Moral Hazard in Health Insurance: What We Know and How We Know It.” Journal of the European Economic Association 16 (4): 957–982. DOI: 10.1093/jeea/jvy017.
Eurostat. n.d. Household Budget Survey: Microdata. European Commission. Accessed November 17, 2026.
Gehrke, E. I., and Y. Jin. 2026. Income and Price Elasticities of Demand for Food in the EU. Research report. Zenodo. DOI: 10.5281/zenodo.19553632.
Havranek, Tomas, Zuzana Irsova, and Karel Janda. 2012. “Demand for Gasoline Is More Price-Inelastic Than Commonly Thought.” Energy Economics 34 (1): 201–207. DOI: 10.1016/j.eneco.2011.09.003.
Houthakker, Hendrik S. 1957. “An International Comparison of Household Expenditure Patterns, Commemorating the Centenary of Engel’s Law.” Econometrica 25 (4): 532–551.
Hughes, Jonathan E., Christopher R. Knittel, and Daniel Sperling. 2008. “Evidence of a Shift in the Short-Run Price Elasticity of Gasoline Demand.” Energy Journal 29 (1): 113–134. DOI: 10.5547/ISSN0195-6574-EJ-Vol29-No1-9.
Jensen, Robert T., and Nolan H. Miller. 2008. “Giffen Behavior and Subsistence Consumption.” American Economic Review 98 (4): 1553–1577. DOI: 10.1257/aer.98.4.1553.
Levin, Laurence, Matthew S. Lewis, and Frank A. Wolak. 2017. “High Frequency Evidence on the Demand for Gasoline.” American Economic Journal: Economic Policy 9 (3): 314–347. DOI: 10.1257/pol.20140093.
Marshall, Alfred. 1890. Principles of Economics. Book III, chapter IV, “The Elasticity of Wants.” Online edition, Library of Economics and Liberty.
Muhammad, Andrew, James L. Seale Jr., Birgit Meade, and Anita Regmi. 2011. International Evidence on Food Consumption Patterns: An Update Using 2005 International Comparison Program Data. Technical Bulletin 1929. Washington, DC: USDA Economic Research Service.
Nsabimana, Aimable. 2024. “Analysis of Household Demand Patterns Using Household Data: Re-thinking the Use of Unit Values or Community Prices.” WIDER Working Paper 2024/22. Helsinki: UNU-WIDER. DOI: 10.35188/UNU-WIDER/2024/480-9.
Nzayiramya, Savant, Andrew Muhammad, and Anthony Baffoe-Bonnie. 2025a. “A Global Assessment of Food and Non-Food Spending: Evidence from 173 Countries and Implications for Food Security.” Agriculture & Food Security 14: 20. DOI: 10.1186/s40066-025-00538-z.
Nzayiramya, Savant, Andrew Muhammad, and Anthony Baffoe-Bonnie. 2025b. “Correction: A Global Assessment of Food and Non-Food Spending: Evidence from 173 Countries and Implications for Food Security.” Agriculture & Food Security 14: 40. DOI: 10.1186/s40066-025-00567-8.
Samuelson, Paul A. 1938. “A Note on the Pure Theory of Consumer’s Behaviour.” Economica 5 (17): 61–71. DOI: 10.2307/2548836. Linked full-text reproduction.
Slutsky, Eugen. 1915. “Sulla teoria del bilancio del consumatore.” Giornale degli Economisti e Rivista di Statistica 51: 1–26. English translation by Olga Ragusa, “On the Theory of the Budget of the Consumer,” reproduced in Readings in Price Theory; linked reprint preview.
Tafere, Kibrom, Zelekawork Paulos, Alemayehu Seyoum Taffesse, Negussie Tefera, and Seneshaw Tamru. 2010. Food Demand Elasticities in Ethiopia: Estimates Using Household Income Consumption Expenditure (HICE) Survey Data. ESSP II Working Paper 11. IFPRI/Ethiopia Strategy Support Program.
U.S. Bureau of Labor Statistics. n.d. Consumer Expenditure Surveys Public Use Microdata: Getting Started Guide. Accessed November 17, 2026.
USDA Economic Research Service. n.d. International Food Consumption Patterns: Documentation. Accessed November 17, 2026.
Varian, Hal R. 1982. “The Nonparametric Approach to Demand Analysis.” Econometrica 50 (4): 945–973.
World Bank. 2024. “Global Purchasing Power Parities Data Released for 2021.” May 30.
World Bank. n.d. Living Standards Measurement Study: Consumption and Welfare. Accessed November 17, 2026.
Zereyesus, Yacob Abrehe, Tian Xia, Noé J. Nava, Xianghong Li, and Lila Cardell. 2025. International Income and Price Elasticity Estimates: An Update. USDA Economic Research Service, Technical Bulletin 1971. DOI: 10.32747/2025.9228006.ers.