Engel Curve: How Spending Patterns Shift with Income

The Engel curve traces the empirical relationship between household income and the quantity demanded of any good, revealing that food’s budget share systematically falls as income rises while luxuries claim a growing share. Formalised by Ernst Engel in 1857, this framework provides the foundational logic for understanding how consumption patterns evolve across the income distribution. When Engel studied working-class household budgets in Saxony, he found that poorer families spent over 60 percent of their income on food while richer families spent less than 30 percent, even though absolute food expenditure rose with income. This regularity, now called Engel’s Law, is the most replicated stylised fact in applied microeconomics. The broader concept plots how the quantity or expenditure share of a good varies with total household resources, holding prices constant. Different goods exhibit different shapes: necessities bend toward the income axis, luxuries bend toward the expenditure axis, and inferior goods slope downward. Consumer theory and revealed preference frameworks rely on these relationships to map preferences from observed choices.

What the Engel Curve Shows

Ernst Engel was a 19th-century Saxon statistician who pioneered the systematic study of household budgets. By examining expenditure patterns across different income groups, he identified a regularity that has survived nearly 170 years of empirical scrutiny: as household income increases, the proportion of income spent on food declines. This does not mean that richer households eat less; absolute spending on food rises with income, but it rises more slowly than income itself. The food budget share falls.

This observation generalises beyond food. Every good or service has its own relationship with income, captured by its Engel curve. Necessities, such as basic housing, staple foods, and utilities, exhibit falling budget shares as income rises. Luxuries, such as international travel, fine dining, and high-end electronics, claim a growing share of the budget as income increases. Inferior goods, such as instant noodles or bus travel for high-income households, actually see declining absolute consumption as income rises, as households substitute toward higher-quality alternatives.

The Engel curve is fundamental to microeconomics because it connects income, the single most important determinant of consumption, to the demand for individual goods. While price elasticity captures how consumers respond to price changes, income elasticity, the slope of the Engel curve in logarithmic form, captures how they respond to changes in purchasing power. Together, price and income elasticities provide a complete statistical description of demand behaviour.

Modern applications extend far beyond describing consumption patterns. Engel curves underpin poverty measurement, inflation bias estimation, equivalence scale calculation, and the demand systems used in computable general equilibrium models. The consumption function at the macroeconomic level aggregates millions of microeconomic Engel curves, linking aggregate income to aggregate spending across categories.

Advertisement

Engel Curve in Equations

The Engel curve formalises the relationship between income and demand through a set of precisely defined equations. The derivation proceeds from the basic definition to flexible modern specifications.

Basic Definition

For good (i), the Engel curve is the function mapping total expenditure or income to quantity demanded:

$$ q_i = f_i(y, mathbf{p}) $$

where (q_i) is the quantity demanded, (y) is total expenditure or income, and (mathbf{p}) is the vector of prices held constant. The function (f_i) describes how consumption of good (i) changes as resources expand, assuming prices remain fixed.

Income Elasticity of Demand

The income elasticity of demand measures the responsiveness of quantity demanded to changes in income:

$$ eta_i = frac{partial q_i}{partial y} cdot frac{y}{q_i} $$

This elasticity classifies goods into three categories. When (eta_i < 0), the good is inferior, meaning consumption falls as income rises. When (0 < eta_i < 1), the good is a necessity, meaning consumption rises but more slowly than income, causing the budget share to fall. When (eta_i > 1), the good is a luxury, meaning consumption rises faster than income, causing the budget share to increase.

Engel Aggregation Condition

Adding up the budget-share-weighted income elasticities across all goods yields the Engel aggregation condition:

$$ sum_i w_i eta_i = 1 $$

where (w_i = p_i q_i / y) is the budget share of good (i). This condition constrains the income elasticities: if some goods have elasticities greater than one, others must have elasticities less than one. The weighted average income elasticity across all goods equals unity, reflecting the fact that total expenditure equals total income.

Common Functional Forms

Economists have estimated Engel curves using various functional forms, each with different implications for how budget shares change with income.

The linear Engel curve specifies the budget share as a linear function of total expenditure:

$$ w_i = alpha_i + beta_i y $$

This specification is simple but problematic. It implies that budget shares change at a constant rate with income, and for sufficiently high income, some budget shares become negative, which is impossible. This saturation problem limits the usefulness of the linear form for modelling consumption across wide income ranges.

The Working-Leser semi-log specification replaces income with the logarithm of income:

$$ w_i = alpha_i + beta_i log y $$

This specification fits Engel’s Law for food remarkably well. The budget share declines at a decreasing rate as income rises, asymptotically approaching a constant. It avoids the saturation problem of the linear form and provides a good fit for necessities. However, it cannot capture goods that switch between necessity and luxury status across the income distribution.

Quadratic Engel Curve and QUAIDS

Banks, Blundell, and Lewbel (1997) introduced the quadratic Engel curve, adding a squared log-income term:

$$ w_i = alpha_i + beta_i log y + gamma_i (log y)^2 $$

The quadratic term (gamma_i) allows the slope of the Engel curve to change across the income distribution. When (gamma_i) is positive, a good that is a necessity at low income levels can become a luxury at higher income levels. When (gamma_i) is negative, the reverse occurs. This flexibility captures real consumption behaviour that simpler specifications miss. For example, alcohol expenditure behaves as a necessity at very low incomes, a luxury at middle incomes, and levels off at very high incomes. The quadratic specification accommodates this non-monotonic pattern.

The Quadratic Almost Ideal Demand System (QUAIDS) builds on this quadratic Engel curve to create a complete demand system that is consistent with utility maximisation. The QUAIDS model generalises the popular Almost Ideal Demand System (AIDS) by adding the quadratic income term, allowing for more realistic income responses while maintaining theoretical coherence.

Lewbel Rank Theorem

Lewbel (1991) provided a fundamental classification of demand systems through the concept of rank. A demand system has rank (R) if all Engel curves can be written as a sum of (R) functions of income only. Linear systems have rank 2, meaning they can be represented using two income functions. The QUAIDS model has rank 3, which is the maximum possible rank for a demand system that is consistent with utility maximisation. This rank-3 property gives QUAIDS its flexibility: it can capture the quadratic curvature in Engel curves that rank-2 systems cannot, without violating the axioms of consumer preferences.

Key Assumptions and Limitations

The Engel curve framework relies on several key assumptions. First, prices are held constant when tracing the income-consumption relationship. Second, household preferences are homothetically separable in goods so that total expenditure summarises income effects. Third, tastes do not change systematically with income.

Several important limitations challenge these assumptions. First, the aggregation problem, formalised by Lewbel (1991), shows that household-level Engel curves do not generally aggregate to a representative-consumer demand system unless preferences satisfy strict conditions, such as the Gorman or PIGLOG form. Aggregation from micro to macro requires care, and simply averaging individual Engel curves can produce misleading results.

Second, household composition matters critically. A family of five has different consumption needs than a single person at the same income level. Equivalence scales are needed to compare welfare across households of different sizes, and these scales are typically estimated from Engel-curve techniques. The OECD modified equivalence scale, widely used in international comparisons, rests on this logic.

Third, endogeneity plagues Engel curve estimation. Income may be jointly determined with consumption choices. A household that particularly values dining out may work more hours or choose higher-paying jobs to afford restaurant meals, creating a feedback loop between income and consumption that biases simple estimates.

Fourth, quality variation confounds the interpretation of Engel curves. As income rises, households do not simply buy more of the same good; they buy higher-quality versions. A household shifting from instant coffee to freshly roasted beans appears to increase its coffee expenditure, but part of this increase reflects a quality upgrade rather than a pure income effect. Standard Engel curves, which treat all units of a good as identical, miss this distinction.

Fifth, cross-section and time-series Engel curves diverge sharply. Income elasticities depend on the price environment and demographic composition, both of which change over time. A cross-sectional elasticity estimated from a single year’s data may not predict how demand evolves as a country grows richer over decades.

Sixth, the linear specification suffers from a saturation problem, implying eventual negative quantities at very high income levels, which is impossible. Even the Working-Leser form imposes monotonicity that some goods violate. The quadratic specification resolves many of these issues but introduces collinearity between the log and squared-log terms that requires careful econometric treatment.

Empirical Evidence for the Engel Curve

The empirical literature on Engel curves spans over 160 years and confirms Engel’s Law as one of the most robust regularities in economics. Houthakker (1957) provided the first major international replication, using data from 30 countries. He found that the income elasticity of demand for food averaged approximately 0.6, confirming that food is a necessity in every country studied, regardless of cultural differences, economic structure, or development level. This cross-country consistency gave Engel’s Law a status comparable to the law of gravity in the physical sciences.

Banks, Blundell, and Lewbel (1997) examined UK Family Expenditure Survey data spanning 1970 to 1986. Their key finding was that many goods exhibit quadratic Engel curves that the simpler Working-Leser specification cannot capture. Alcohol, for instance, displays a hump-shaped budget share, rising with income at low levels and falling at high levels. Transport and personal services show the opposite pattern. These results validated the QUAIDS model and demonstrated that rank-3 demand systems are empirically necessary, not just theoretically elegant.

Building on Banks, Blundell, and Lewbel (1997), Blundell, Browning, and Crawford (2003), published in Econometrica, used nonparametric methods to estimate Engel curves. Their revealed preference approach provided additional evidence that nonlinear Engel curves — including the quadratic specification — are empirically necessary, while also showing that some goods exhibit even more complex patterns at the extremes of the income distribution.

Lewbel (1991) tested the rank of demand systems directly, finding that UK expenditure data rejects rank 2 in favour of rank 3, further supporting the QUAIDS framework. The rank-3 property means that income effects cannot be captured by two functions of income alone; a third function is needed, which the quadratic term provides.

Almås (2012), in the American Economic Review, used Engel curves to estimate true cross-country price level differences, correcting the Penn effect for substitution bias. The standard Penn effect compares price levels across countries using fixed consumption baskets. Almås showed that as countries grow richer, their consumption baskets shift toward more expensive goods, overstating the true price level difference. By using Engel curves to infer real income from food budget shares, he estimated that standard PPP calculations systematically overstate the incomes of poorer countries, meaning that international income inequality is substantially larger than conventional measures indicate.

Hamilton (2001), also in the AER, used food Engel curves to estimate true CPI inflation. His insight was elegant: if official CPI overstates inflation, then households at a given nominal income are actually richer than the CPI suggests, and their food budget share should be lower than predicted by the historical Engel curve. By comparing observed food shares with those predicted by the historical Engel curve, Hamilton estimated that US CPI overstated inflation by roughly 1 percentage point per year, consistent with the Boskin Commission findings but derived entirely from expenditure data without relying on price indices.

The US Bureau of Labor Statistics Consumer Expenditure Survey and the World Bank Living Standards Measurement Studies continue to provide the microdata that underpins modern Engel curve estimation. The USDA Economic Research Service has documented nonlinear Engel curves for specific food categories, showing that pork and egg consumption follow complex patterns that reflect both income effects and shifting dietary preferences.

Sources: US BLS Consumer Expenditure Survey 2024; UK ONS Family Expenditure Survey; Banks, Blundell & Lewbel (1997).

Good Category	Income Elasticity ((eta))	Classification	Country / Period	Source
Staple food (cereals)	0.20	Necessity	Developing avg., 2000s	FAO/IFPRI
Food at home (composite)	0.55	Necessity	OECD, 2010s	Houthakker; USDA
Clothing	0.85	Near-unitary	UK, 1980–2010	BBL (1997)
Healthcare services	1.10	Mild luxury	OECD avg.	Newhouse (1977)
Restaurants/dining out	1.45	Luxury	UK, 1980s	BBL (1997)
International travel	2.20	Strong luxury	Global, 2010s	UNWTO data
Luxury fashion	2.80	Strong luxury	Global LVMH/Hermès	Industry analyst
Tobacco (high income)	-0.30	Inferior	US, post-1990	BLS CEX

How the Engel Curve Matters

The Engel curve framework extends far beyond academic consumer theory. It shapes how governments measure poverty, estimate inflation, and design social policy.

First, the poverty line setting in the United States rests directly on Engel’s Law. Mollie Orshansky’s 1965 US poverty threshold was calculated as three times the cost of a minimum adequate diet. This multiplier of three was derived from Engel’s Law: because the average poor household spent approximately one-third of its income on food, multiplying the food budget by three approximated the total income needed to meet basic needs. The Orshansky method, still used in modified form today, is Engel’s Law applied to social policy. Every time the US Census Bureau updates the poverty threshold, it relies on a framework built on Engel’s empirical observation from 1857.

Second, Engel curves provide a powerful tool for measuring CPI bias and true inflation. Hamilton (2001) and Costa (2001) exploited the logic that if official CPI overstates inflation, then households at a given nominal income are actually richer than reported, and their food budget share should be lower than the historical Engel curve predicts. By measuring the gap between actual and predicted food shares, they estimated that US CPI overstates inflation by roughly 1 percentage point per year. This finding has profound implications for fiscal policy, social security adjustments, and purchasing power assessments. The Consumer Price Index is the most important economic indicator for monetary and fiscal policy, and Engel curve evidence suggests it may contain systematic upward bias.

Third, equivalence scales for international comparison depend on Engel-curve estimation. Comparing the welfare of a single-person household with a family of four at the same income level requires adjusting for differences in consumption needs. The OECD modified equivalence scale, which assigns a value of 1.0 to the household head, 0.5 to each additional adult, and 0.3 to each child, is estimated using Engel-curve techniques. The Luxembourg Income Study and the World Bank’s PovcalNet database rely on these scales to make meaningful cross-country poverty comparisons. Without Engel-curve-based equivalence scales, global poverty estimates would be far less precise.

Fourth, demand systems used in trade and policy analysis are built on Engel curve foundations. The QUAIDS and AIDS models are standard tools in tax incidence studies, computable general equilibrium models, and food policy analysis at the IMF, World Bank, USDA, and Eurostat. When a government considers raising value-added tax on food, the welfare impact depends on how budget shares vary across income groups, information that Engel curves provide. When the FAO projects future food demand, it uses income elasticities estimated from Engel curves.

Fifth, food security and agricultural economics rely heavily on Engel curves. Projecting future food demand requires knowing how caloric intake and dietary composition change as incomes rise in developing countries. The FAO and the International Food Policy Research Institute (IFPRI) use Engel curves to forecast that global food demand will rise by approximately 60 percent by 2050, driven primarily by income growth in Asia and Africa. These projections inform agricultural investment decisions, land-use planning, and climate adaptation strategies.

Sixth, marketing and consumer research explicitly track Engel curve elasticities. Luxury goods firms such as LVMH and Hermès monitor income elasticities closely. The income elasticity of luxury fashion is estimated at 2.5 to 3.0, meaning that a 1 percent increase in disposable income generates a 2.5 to 3.0 percent increase in luxury spending. This high elasticity makes luxury firms highly sensitive to macroeconomic conditions but also provides outsized growth during economic expansions.

Seventh, inequality measurement uses Engel curve data. Estimates of income inequality, including the Atkinson index and the Piketty-Saez top income shares, rely on household survey data organised by income brackets. Engel curves help researchers adjust for underreporting of income at the top of the distribution by using expenditure patterns, which are typically reported more accurately than income, to infer true resources. The distributional effects of inflation also depend on how budget shares differ across income groups.

Eighth, modern microfinance and behavioural studies use food Engel curves as low-frequency welfare indicators. In randomised controlled trials conducted by Banerjee, Duflo, and others, researchers use the food budget share as a proxy for household welfare when direct income measures are unreliable. A declining food share indicates rising real income, even when nominal income is difficult to measure. This application extends Engel’s original insight to 21st-century development economics.

Ninth, emerging market analysis relies on Engel curve dynamics. The rise of “premiumisation” in China, India, and Indonesia is fundamentally an Engel curve story. As median household income crosses certain thresholds, the share of premium consumer products rises non-linearly. Middle-class households in these countries are shifting from necessities to luxuries at a pace predicted by income elasticities. This premiumisation drives investment decisions in consumer goods industries and shapes trade patterns.

Tenth, climate policy uses Engel curves to study whether environmental services are luxuries or necessities. Research by Stern and others finds that clean air, biodiversity conservation, and climate mitigation have income elasticities greater than one, meaning they are luxury goods. This finding implies that as countries grow richer, their willingness to pay for environmental protection increases faster than income, suggesting that economic growth may eventually generate demand for climate action. Luxury goods and permanent income dynamics both play into this analysis.

MASEconomics Explains

Conclusion

Engel curve analysis remains the foundational tool for understanding how spending patterns shift with income across households and countries. Ernst Engel’s 1857 stylised fact about food spending remains the most replicated regularity in applied economics, and the framework that bears his name underpins poverty measurement, CPI bias estimation, equivalence scales, demand system estimation, and forecasts of long-run consumption patterns. Modern flexible specifications like QUAIDS allow goods to switch between necessity and luxury status as income changes, capturing the non-monotonic behaviour that simpler models miss. From setting the US poverty line to projecting global food demand, the Engel curve continues to shape economic policy and research.

Did you find this article helpful? Share it with someone who loves economics. And remember, at MASEconomics, we make complex ideas simple.

Advertisement