52.01.02 · economics / microeconomics

Revealed preference and Afriat's theorem: cyclical consistency and the rationalizability of demand

shipped3 tiersLean: none

Anchor (Master): Samuelson 1938 A note on the pure theory of consumer's behaviour (Economica 5); Afriat 1967 The construction of utility functions from expenditure data (QJE 80); Diewert 1973 Afriat and revealed preference theory (Review of Economic Studies 40); Varian 1982 The nonparametric approach to demand analysis (Econometrica 50)

Intuition Beginner

A consumer walks into a market with a fixed budget. Prices and the budget determine which bundles of goods are affordable, and the consumer picks one. Do this many times, with different prices and budgets, and a question arises: is the consumer's behaviour consistent with maximising some stable preference over bundles? They need not tell us their preferences; their choices are the data. The question is whether those choices cohere.

The clearest test is the Weak Axiom of Revealed Preference. If at one price-budget the consumer chose bundle $A$ when bundle $B$ was also affordable, then $A$ is "revealed preferred" to $B$ . If, at a later price-budget, the consumer chooses $B$ when $A$ is affordable, we have a contradiction: $A$ beat $B$ once, and $B$ beat $A$ the other time. No single stable preference can produce both. The Weak Axiom forbids this two-step contradiction.

The deeper question is whether longer chains are also forbidden. Suppose $A$ is revealed preferred to $B$ , $B$ to $C$ , and then $C$ to $A$ . That is a cycle. Afriat's theorem (1967) is the profound result: the absence of any such cycle is not only necessary but also sufficient. If the data are cyclically consistent — there are no preference cycles of any length — then a well-behaved utility function exists that rationalises every choice. The consumer behaves as if maximising. We never needed to see the utility function; the absence of cycles proves one exists.

Visual Beginner

The picture shows a two-good consumption space with three price-budget lines and three chosen bundles $A$ , $B$ , $C$ . Each chosen bundle sits on its budget line, with the shaded triangle below the line marking the affordable set. Arrows trace the revealed-preference relation: $A$ is in $B$ 's affordable set, so $B$ is revealed preferred to $A$ ; similarly for the other pairs. In the left panel the arrows form a cycle (a GARP violation); in the right panel they do not (a GARP-consistent data set).

The acyclic picture on the right is the test: when no cycle of any length exists in the revealed-preference relation, Afriat's theorem guarantees a utility function exists that explains every choice.

Worked example Beginner

Test whether a three-observation demand data set satisfies the Weak Axiom of Revealed Preference.

Step 1. Set up three observations of a consumer's choice over two goods (good 1, good 2). At observation 1, prices are $p^{1} = (1, 2)$ and the chosen bundle is $x^{1} = (4, 2)$ . At observation 2, prices are $p^{2} = (2, 1)$ and the chosen bundle is $x^{2} = (2, 4)$ . At observation 3, prices are $p^{3} = (2, 2)$ and the chosen bundle is $x^{3} = (3, 3)$ .

Step 2. Compute each budget expenditure. The expenditure at observation $i$ is $p^{i} \cdot x^{i}$ . For observation 1, $1 \times 4 + 2 \times 2 = 8$ . For observation 2, $2 \times 2 + 1 \times 4 = 8$ . For observation 3, $2 \times 3 + 2 \times 3 = 12$ .

Step 3. Test direct revealed preference. Bundle $x^{j}$ is directly revealed preferred to $x^{i}$ if at prices $p^{j}$ bundle $x^{i}$ was affordable ( $p^{j} \cdot x^{i} \leq p^{j} \cdot x^{j}$ ) but a different bundle $x^{j}$ was chosen. Compute: $p^{2} \cdot x^{1} = 2 \times 4 + 1 \times 2 = 10$ , which exceeds the expenditure $p^{2} \cdot x^{2} = 8$ . So $x^{1}$ was not affordable at observation 2 — observation 2 does not reveal-prefer $x^{2}$ to $x^{1}$ . Similarly $p^{1} \cdot x^{2} = 1 \times 2 + 2 \times 4 = 10 > 8$ , so $x^{2}$ was not affordable at observation 1.

Step 4. Cycle check. With neither observation revealing a preference over the other's bundle in either direction, there is no direct preference relation and hence no cycle. The data satisfy the Weak Axiom.

What this tells us: the Weak Axiom is checked by computing pairwise expenditures and verifying that whenever one bundle is affordable at the other's chosen prices, the reverse is not also true. Three observations is the smallest substantive case for a cycle, and the check extends to chains of any length.

Check your understanding Beginner

Exercise (easy, multiple choice).

The Weak Axiom of Revealed Preference (WARP) states that:

A. Consumers always prefer more goods to fewer B. If bundle $A$ is directly revealed preferred to bundle $B$ , then $B$ cannot be directly revealed preferred to $A$ C. Consumers maximise a unique utility function D. Demand curves always slope downward

Hint

WARP forbids a two-step contradiction in revealed preferences. It does not address cycles of length three or more.

Answer

B. WARP states that if $A$ is directly revealed preferred to $B$ (meaning $A$ was chosen when $B$ was affordable), then $B$ cannot also be directly revealed preferred to $A$ . The axiom forbids the two-step preference contradiction but does not by itself rule out longer cycles; that requires the Strong Axiom (SARP) or the Generalised Axiom (GARP).

Exercise (easy, true-false).

Afriat's theorem states that a finite demand data set can be rationalised by a well-behaved (continuous, increasing, concave) utility function if and only if the data satisfy cyclical consistency (GARP).

Hint

This is the precise content of Afriat's 1967 theorem: cyclical consistency is both necessary and sufficient for utility-rationalisability, and the utility can be constructed explicitly.

Answer

True. Afriat's theorem establishes the equivalence between cyclical consistency (no cycles of any length in the revealed-preference relation, a condition known as GARP) and the existence of a well-behaved utility function that rationalises the data. The theorem is constructive: it gives explicit inequalities (the Afriat inequalities) whose solution yields the utility function. The theorem is profound because it shows that observable choice data alone, without any cardinal measurement of utility, suffice to recover an ordinal utility function.

Exercise (easy, multiple choice).

What does it mean to say that revealed-preference theory is nonparametric?

A. It assumes a specific functional form for utility (e.g. Cobb-Douglas) B. It tests whether choices are consistent with some utility function, without committing to any particular functional form C. It uses no prices or quantities D. It applies only to markets with no government intervention

Hint

"Nonparametric" in this context means no functional-form assumption on the utility function. The test asks whether any well-behaved utility could rationalise the data.

Answer

B. Revealed-preference theory is nonparametric in the sense that it does not assume a specific functional form for the utility function (Cobb-Douglas, CES, linear, etc.). Instead it asks whether any well-behaved utility function exists that rationalises the observed choices. If the data satisfy GARP, Afriat's theorem guarantees such a function exists, without restricting its functional form. This makes revealed-preference testing much more general than parametric demand estimation.

Formal definition Intermediate+

Revealed-preference theory characterises which finite demand data sets are consistent with utility maximisation. Its core notions are the direct and indirect revealed-preference relations, the Weak, Strong, and Generalised Axioms, and the Afriat inequalities.

Definition (demand data set). A demand data set is a finite collection of pairs $(p^{i}, x^{i})_{i = 1}^{N}$ where $p^{i} \in R_{++}^{n}$ is a price vector and $x^{i} \in R_{+}^{n}$ is the bundle chosen at those prices under the budget $p^{i} \cdot x \leq p^{i} \cdot x^{i}$ .

Definition (direct revealed preference). Bundle $x^{i}$ is directly revealed preferred to bundle $x^{j}$ , written $x^{i} ⪰_{R} x^{j}$ , if $p^{i} \cdot x^{j} \leq p^{i} \cdot x^{i}$ (bundle $x^{j}$ was affordable at observation $i$ ). The relation is strict, written $x^{i} ≻_{R} x^{j}$ , if $p^{i} \cdot x^{j} < p^{i} \cdot x^{i}$ .

Definition (GARP). The data set satisfies the Generalised Axiom of Revealed Preference if there is no cycle of direct-revealed-preference relations in which at least one step is strict. Equivalently, if $x^{i_{1}} ⪰_{R} x^{i_{2}} ⪰_{R} \dots ⪰_{R} x^{i_{k}} ⪰_{R} x^{i_{1}}$ , then none of the steps is strict (all the bundles cost exactly the same at their respective prices).

Definition (Afriat inequalities). Given a data set $(p^{i}, x^{i})_{i = 1}^{N}$ , the Afriat inequalities are the system in unknowns $(λ^{i} > 0, U^{i})_{i = 1}^{N}$ :

U^{j} \leq U^{i} + λ^{i} p^{i} \cdot (x^{j} - x^{i}) for all i, j .

Counterexamples to common slips Intermediate+

Conflating WARP and SARP. WARP forbids only two-step preference contradictions; SARP (Houthakker 1950) forbids cycles of any length in the strict revealed-preference relation. For finite data sets with distinct bundles, SARP and GARP are essentially equivalent, but WARP is weaker and does not by itself imply rationalisability.
Treating revealed preference as a behavioural theory. Revealed preference is an analytical framework for testing consistency with utility maximisation, not a claim about how consumers actually think. A consumer who violates GARP is not necessarily irrational in any deep sense; the violation only shows that no single utility function explains the choices.
Assuming GARP guarantees a unique utility. GARP guarantees existence of a rationalising utility, not uniqueness. Many utility functions (positive monotone transformations of each other, and others besides) rationalise the same data.

Economic theory Intermediate+

Theorem (Afriat 1967). For a finite demand data set $(p^{i}, x^{i})_{i = 1}^{N}$ , the following are equivalent:

The data satisfy GARP (cyclical consistency).
There exist numbers $λ^{i} > 0$ and $U^{i}$ ( $i = 1, \dots, N$ ) satisfying the Afriat inequalities.
There exists a continuous, increasing, concave utility function $u : R_{+}^{n} \to R$ such that $x^{i}$ maximises $u$ on the budget set ${x : p^{i} \cdot x \leq p^{i} \cdot x^{i}}$ for every $i$ .

Reconstruction. The theorem's three-way equivalence establishes that observable choice data alone — without any cardinal measurement of utility — suffice to recover an ordinal utility function, and that cyclical consistency is the precise test. The implication (3) $\Rightarrow$ (1) is straightforward: if a utility function rationalises the data, then any revealed-preference cycle would force equal utility on all bundles in the cycle, contradicting the strictness. The implication (1) $\Rightarrow$ (2) is the linear-programming core of Afriat's proof: cyclical consistency ensures the Afriat inequalities have a feasible solution. The implication (2) $\Rightarrow$ (3) constructs $u$ explicitly as a piecewise-linear concave function whose supporting hyperplanes are given by the $λ^{i}$ and $U^{i}$ :

u (x) = i = 1, \dots, N min {U^{i} + λ^{i} p^{i} \cdot (x - x^{i})} .

This $u$ is concave (minimum of affine functions), increasing in $x$ (since $λ^{i} > 0$ and $p^{i} > 0$ ), and satisfies $u (x^{i}) = U^{i}$ with $x^{i}$ maximising $u$ on the $i$ -th budget set.

Bridge. Afriat's theorem builds toward 52.04.02 signaling games by showing how observable behaviour can reveal an underlying preference structure without any direct measurement of utility, and appears again in 52.01.01 microeconomics foundations as the rigorous foundation for the demand curve that the foundations unit takes as given. The foundational reason the theorem is profound is that it closes the gap between observability (choices, prices) and unobservability (utility): the absence of preference cycles is not merely a necessary condition but also a sufficient one, so the testable consistency condition fully characterises rationalisability. This is exactly the structure that identifies a behavioural dataset with the existence of a rationalising preference, and the bridge is from finite observed choices to a constructed utility function, with the Afriat inequalities as the constructive intermediary. The pattern generalises across the modern nonparametric programme — Varian's GARP test, the collective-household model of Chiappori, the characterisation of production efficiency — each of which treats consistency of observed data as the necessary and sufficient condition for an underlying structural representation, and the central insight is that ordinal preferences are fully pinned down by choice behaviour under the single testable axiom of cyclical consistency.

Exercises Intermediate+

Exercise 2 (easy, multiple choice).

A consumer's data violate WARP. What can we conclude?

A. The consumer has no preferences at all B. No single utility function can rationalise the data; the consumer's choices are inconsistent with utility maximisation C. The consumer must be price-insensitive D. The data are insufficient to draw any conclusion

Hint

A WARP violation is a direct two-step preference contradiction. No single utility function can produce a contradiction.

Answer

B. A WARP violation means $A$ was chosen when $B$ was affordable, and $B$ was chosen when $A$ was affordable (with at least one case being strict). No single utility function can produce this: if $u (A) > u (B)$ , the second choice is irrational; if $u (B) > u (A)$ , the first is irrational; if $u (A) = u (B)$ , neither is strictly preferred and at least one strict reveal is contradicted. The data are inconsistent with utility maximisation.

Exercise 3 (medium, symbolic).

State the Afriat inequalities for a two-observation data set and show that their feasibility is equivalent to WARP in this case.

Hint

For two observations the Afriat inequalities reduce to four constraints. Use the freedom in $λ^{i}$ to match the WARP condition.

Answer

For two observations $(p^{1}, x^{1}), (p^{2}, x^{2})$ the Afriat inequalities in $(λ^{1}, λ^{2}, U^{1}, U^{2})$ are:

$U^{1} \leq U^{1} + λ^{1} p^{1} \cdot (x^{1} - x^{1}) = U^{1}$ (satisfied by construction),
$U^{2} \leq U^{1} + λ^{1} p^{1} \cdot (x^{2} - x^{1})$ ,
$U^{1} \leq U^{2} + λ^{2} p^{2} \cdot (x^{1} - x^{2})$ ,
$U^{2} \leq U^{2}$ (satisfied by construction).

The two substantive inequalities combine to $λ^{1} p^{1} \cdot (x^{2} - x^{1}) + λ^{2} p^{2} \cdot (x^{1} - x^{2}) \geq 0$ . If both $x^{2}$ was affordable at observation 1 ( $p^{1} \cdot x^{2} \leq p^{1} \cdot x^{1}$ ) and $x^{1}$ was affordable at observation 2 ( $p^{2} \cdot x^{1} \leq p^{2} \cdot x^{2}$ ), with at least one strict, the left side is negative for all positive $λ$ , contradicting the inequality — a WARP violation. Conversely, if WARP holds, the inequalities can be satisfied by choosing $λ$ appropriately. So for two observations the Afriat inequalities are feasible if and only if WARP holds.

Exercise 5 (medium, multiple choice).

The integrability problem of demand theory asks:

A. Whether demand curves can be integrated to recover the consumer's utility function B. Whether consumers integrate information from multiple sources C. Whether firms can integrate vertically D. Whether market demand equals the sum of individual demands

Hint

The integrability problem is the recovery of an indirect utility function from observed demand derivatives, subject to symmetry and negative semi-definiteness of the Slutsky matrix.

Answer

A. The integrability problem asks: given an observed demand function $x (p, w)$ , does there exist an indirect utility function that, when maximised, generates that demand? The answer requires the Slutsky matrix of substitution effects to be symmetric and negative semi-definite — these conditions are equivalent to the existence of a potential function (the indirect utility) whose derivatives give the demand. The integrability problem is the differential counterpart of the finite-data Afriat theorem: both ask whether observed behaviour is rationalisable.

Exercise 6 (medium, multiple choice).

Varian's nonparametric approach to demand analysis is "nonparametric" because:

A. It uses no price data B. It assumes Cobb-Douglas utility but estimates its parameters C. It tests GARP directly on observed data without assuming a parametric family for utility D. It applies only to non-durable goods

Hint

Nonparametric means no functional-form restriction on the utility function. GARP is the testable consistency condition that does not require any parametric specification.

Answer

C. Varian's nonparametric approach tests GARP directly on a finite demand data set, without assuming any parametric family (Cobb-Douglas, CES, translog, etc.) for the utility function. The test asks whether some well-behaved utility could rationalise the data; if GARP holds, Afriat's theorem guarantees existence. The approach is much more general than parametric demand estimation and is the standard nonparametric tool for testing consumer rationality.

Exercise 7 (hard, symbolic).

Prove that the utility function $u (x) = min_{i = 1, \dots, N} {U^{i} + λ^{i} p^{i} \cdot (x - x^{i})}$ constructed from a solution of the Afriat inequalities is concave, increasing, and rationalises the data (i.e., $x^{i}$ maximises $u$ on the $i$ -th budget set).

Hint

For concavity: $u$ is the pointwise minimum of affine functions of $x$ , which is always concave. For increasing: each affine piece has slope $λ^{i} p^{i} > 0$ . For rationalisation: use the Afriat inequality at observation $i$ to show $u (x^{i}) = U^{i}$ and $u (x) \leq U^{i}$ for any $x$ with $p^{i} \cdot x \leq p^{i} \cdot x^{i}$ .

Answer

Concavity: $u$ is the pointwise minimum of affine functions $f_{i} (x) = U^{i} + λ^{i} p^{i} \cdot (x - x^{i})$ , and the pointwise minimum of affine functions is concave (its hypograph is the intersection of half-spaces, hence convex). Increasing: each $f_{i}$ has gradient $λ^{i} p^{i} \in R_{++}^{n}$ (positive in every component), so $f_{i}$ is increasing in $x$ ; the minimum of increasing functions is increasing. Rationalisation: at observation $i$ , the Afriat inequality with $j = i$ gives $U^{i} \leq U^{i} + λ^{i} p^{i} \cdot (x^{i} - x^{i}) = U^{i}$ , so $u (x^{i}) \leq U^{i}$ . For any $j$ , the Afriat inequality gives $U^{i} \leq U^{j} + λ^{j} p^{j} \cdot (x^{i} - x^{j})$ , which is one of the affine pieces evaluated at $x^{i}$ ; so $u (x^{i}) = U^{i}$ . Now for any $x$ in the $i$ -th budget set ( $p^{i} \cdot x \leq p^{i} \cdot x^{i}$ ), the $i$ -th affine piece gives $f_{i} (x) = U^{i} + λ^{i} p^{i} \cdot (x - x^{i}) \leq U^{i} = u (x^{i})$ . Since $u (x) \leq f_{i} (x) \leq u (x^{i})$ , $x^{i}$ maximises $u$ on the $i$ -th budget set. $□$

Exercise 8 (hard, multiple choice).

Why is Afriat's theorem considered a foundational result in microeconomics?

A. It proves that all consumers are rational B. It establishes the equivalence between an observable consistency condition (GARP) and the existence of an unobservable structural object (a utility function), grounding utility theory entirely in choice data C. It shows that Cobb-Douglas is the correct functional form for utility D. It demonstrates that demand curves always slope upward

Hint

The theorem's significance is epistemological: it shows utility functions can be recovered from choice data alone, without any direct measurement of utility or cardinal comparison across people.

Answer

B. Afriat's theorem establishes the equivalence between GARP (an observable, testable condition on choice data) and the existence of a well-behaved utility function (an unobservable structural object). The theorem grounds utility theory entirely in choice data: we never need to measure utility directly; we only need to observe choices and check cyclical consistency. The result is foundational because it operationalises the ordinalist revolution (Hicks-Allen) that utility is a representation of preference, not a measurable substance, and it makes the test of rationality fully nonparametric.

Advanced results Master

Result 1 (Samuelson's revealed preference, 1938). Paul Samuelson introduced the revealed-preference approach as a behavioural foundation for demand theory, replacing the cardinal utility of the late-nineteenth-century marginalists with choice-based inference. The Weak Axiom of Revealed Preference (WARP) is Samuelson's original consistency condition. The programme reframes utility as a representation of consistent choice rather than as a measurable psychological magnitude, a methodological move that defined ordinalist microeconomics.

Result 2 (Houthakker's SARP, 1950). Hendrik Houthakker extended WARP to the Strong Axiom of Revealed Preference (SARP), which forbids cycles of any length in the strict revealed-preference relation. Houthakker showed that SARP is sufficient for the existence of a rationalising utility function in the case of smooth demand. The move from WARP to SARP closes the gap between the two-step axiom and full rationalisability for finite data.

Result 3 (Afriat's theorem, 1967). Sydney Afriat established the three-way equivalence between GARP (cyclical consistency on finite data), the feasibility of the Afriat inequalities, and the existence of a continuous, increasing, concave utility function that rationalises the data. The theorem is constructive: the Afriat inequalities, when feasible, yield explicit supporting hyperplanes from which the utility function is recovered as a piecewise-linear concave envelope. The theorem is the central result of finite-data revealed-preference theory.

Result 4 (Diewert's linear-programming proof, 1973). Erwin Diewert gave the modern proof of Afriat's theorem using Farkas' lemma and linear-programming duality, replacing Afriat's original construction with a cleaner argument that exposes the equivalence between GARP and the feasibility of the Afriat inequalities. The Diewert proof is now standard in graduate microeconomics.

Result 5 (Varian's GARP test, 1982). Hal Varian operationalised the GARP test as a shortest-path algorithm on the revealed-preference graph: build a directed graph with an edge $i \to j$ whenever $x^{i}$ is revealed preferred to $x^{j}$ , and test whether any cycle contains a strict edge. The algorithm runs in polynomial time and is the standard nonparametric test of consumer rationality. Varian's paper established the nonparametric revealed-preference programme as a practical empirical tool.

Result 6 (Chiappori's collective household model, 1992). Pierre-André Chiappori extended revealed-preference theory to multi-person households, characterising which household demand data are consistent with Pareto-efficient collective decision-making among household members. The characterisation is the analogue of Afriat's theorem for a setting with multiple utility functions aggregated by a Pareto weight, and is the foundation of modern household economics.

Synthesis. Afriat's theorem builds toward 52.04.02 signaling games by showing how observable behaviour can reveal an underlying structural object without direct measurement, and appears again in 52.01.01 microeconomics foundations as the rigorous foundation for the demand theory that the foundations unit treats informally. The foundational reason the theorem is profound is that it closes the epistemological gap between choice data (observable) and utility (unobservable), making the testable consistency condition the full characterisation of rationalisability. This is exactly the structure that identifies a behavioural dataset with the existence of a rationalising preference, and putting these together with the Diewert linear-programming proof and the Varian shortest-path test, the bridge is from finite observed choices to a constructed utility function, with the Afriat inequalities as the constructive intermediary. The pattern generalises across the modern nonparametric programme — Varian's GARP test, Chiappori's collective household, the Brown-Matzkin stochastic-revealed-preference model — each of which treats consistency of observed data as the necessary and sufficient condition for an underlying structural representation, and the central insight is that ordinal preferences are fully pinned down by choice behaviour under the single testable axiom of cyclical consistency.

Full proof set Master

Proposition (Necessity: rationalisability implies GARP). If a continuous, increasing utility function $u$ rationalises the data $(p^{i}, x^{i})_{i = 1}^{N}$ (i.e., $x^{i}$ maximises $u$ on the $i$ -th budget set for every $i$ ), then the data satisfy GARP.

Proof. Suppose for contradiction there is a revealed-preference cycle with at least one strict step: $x^{i_{1}} ⪰_{R} x^{i_{2}} ⪰_{R} \dots ⪰_{R} x^{i_{k}} ⪰_{R} x^{i_{1}}$ , with $x^{i_{m}} ≻_{R} x^{i_{m + 1}}$ strict for some $m$ . Rationalisability gives $u (x^{i_{1}}) \geq u (x^{i_{2}}) \geq \dots \geq u (x^{i_{k}}) \geq u (x^{i_{1}})$ , hence $u$ is constant on the cycle. But the strict step $x^{i_{m}} ≻_{R} x^{i_{m + 1}}$ means $x^{i_{m + 1}}$ was strictly affordable at observation $i_{m}$ (i.e., $p^{i_{m}} \cdot x^{i_{m + 1}} < p^{i_{m}} \cdot x^{i_{m}}$ ), and $x^{i_{m}}$ was chosen; strict optimality gives $u (x^{i_{m}}) > u (x^{i_{m + 1}})$ , contradicting constancy. So no strict cycle exists. $□$

Proposition (Sufficiency: GARP implies rationalisability, via Afriat inequalities). If the data satisfy GARP, then there exist $λ^{i} > 0$ and $U^{i}$ satisfying the Afriat inequalities, and the utility $u (x) = min_{i} {U^{i} + λ^{i} p^{i} \cdot (x - x^{i})}$ rationalises the data.

Proof sketch (Diewert 1973). The Afriat inequalities are feasible if and only if there is no non-negative linear combination that yields a contradiction; by Farkas' lemma, infeasibility is equivalent to the existence of a "bad" linear combination. The bad combination, when unravelled, corresponds exactly to a revealed-preference cycle with at least one strict step (a GARP violation). So GARP implies the Afriat inequalities are feasible, giving $(λ^{i}, U^{i})$ . The piecewise-linear construction $u$ is then concave (minimum of affine), increasing (positive gradients), and rationalises each observation by the argument in Exercise 7. $□$

Connections Master

Microeconomics — scarcity, choice, and equilibrium 52.01.01. Revealed-preference theory is the rigorous foundation for the demand curve that the foundations unit treats informally. The move from cardinal utility to choice-based inference is the methodological pivot of ordinalist microeconomics.
Signaling games and Bayesian Nash equilibrium 52.04.02. Both revealed-preference theory and signaling-game theory extract unobservable structural objects (preferences, types) from observable behaviour (choices, signals). The shared methodology is the inference of the unobservable from the consistency of the observable.
Game theory — Nash equilibrium and strategic interaction 52.04.01. Revealed-preference ideas extend to game-theoretic settings, where the "choices" are strategy profiles and the consistency condition is equilibrium. The Nash equilibrium itself can be characterised as a revealed-preference fixed point of best responses.

Historical & philosophical context Master

Paul Samuelson introduced revealed-preference theory in 1938 ^{[Samuelson1938]} in Economica as a behavioural foundation for demand theory, in conscious opposition to the cardinal utility of the late-nineteenth-century marginalists (Jevons, Menger, Walras). Samuelson's project was to rid microeconomics of unmeasurable psychological magnitudes and to ground demand in observable choice. The Weak Axiom was his original consistency condition.

Hendrik Houthakker extended the approach to the Strong Axiom in 1950 ^{[Houthakker1950]} in Economica, showing that SARP (the absence of revealed-preference cycles of any length) is sufficient for rationalisability under smooth demand. Sydney Afriat's 1967 paper ^[Afriat1967] in the Quarterly Journal of Economics established the full three-way equivalence for finite data and introduced the Afriat inequalities as the constructive machinery. Erwin Diewert's 1973 paper ^{[Diewert1973]} in the Review of Economic Studies gave the modern linear-programming proof via Farkas' lemma. Hal Varian's 1982 Econometrica paper ^[Varian1982] operationalised the GARP test as a polynomial-time shortest-path algorithm and founded the nonparametric revealed-preference programme.

The deeper lineage runs through the ordinalist revolution of Hicks and Allen (1934), which replaced cardinal utility with indifference curves and established that demand theory requires only ordinal preference information. Revealed-preference theory completed the ordinalist project by showing that ordinal preferences are fully pinned down by choice behaviour under the single testable axiom of cyclical consistency. The modern extensions to collective households (Chiappori), to stochastic settings (Brown-Matzkin, McFadden-Richter), and to production theory continue the programme.

Bibliography Master

@article{Samuelson1938,
  author = {Samuelson, P. A.},
  title = {A note on the pure theory of consumer's behaviour},
  journal = {Economica},
  volume = {5},
  number = {17},
  pages = {61--71},
  year = {1938},
}

@article{Houthakker1950,
  author = {Houthakker, H. S.},
  title = {Revealed preference and the utility function},
  journal = {Economica},
  volume = {17},
  number = {66},
  pages = {159--174},
  year = {1950},
}

@article{Afriat1967,
  author = {Afriat, S. N.},
  title = {The construction of utility functions from expenditure data},
  journal = {Quarterly Journal of Economics},
  volume = {80},
  number = {4},
  pages = {643--664},
  year = {1967},
}

@article{Diewert1973,
  author = {Diewert, W. E.},
  title = {Afriat and revealed preference theory},
  journal = {Review of Economic Studies},
  volume = {40},
  number = {3},
  pages = {419--425},
  year = {1973},
}

@article{Varian1982,
  author = {Varian, H. R.},
  title = {The nonparametric approach to demand analysis},
  journal = {Econometrica},
  volume = {50},
  number = {4},
  pages = {945--973},
  year = {1982},
}

@book{MasColellWhinstonGreen1995,
  author = {Mas-Colell, A. and Whinston, M. D. and Green, J. R.},
  title = {Microeconomic Theory},
  publisher = {Oxford University Press},
  address = {New York},
  year = {1995},
}

Prerequisites

52.01.01

Tier anchors

beginner: Varian — Intermediate Microeconomics (Norton 2014) Ch. 6 on demand and the basics of revealed preference; Pollak — Samuelson's revealed preference theory (Journal of Economic Literature 2018)
intermediate: Mas-Colell-Whinston-Green — Microeconomic Theory (Oxford 1995) Ch. 2-3 on demand theory and revealed preference; Varian — Microeconomic Analysis (Norton 1992) Ch. 8 on the integrability problem
master: Samuelson 1938 A note on the pure theory of consumer's behaviour (Economica 5); Afriat 1967 The construction of utility functions from expenditure data (QJE 80); Diewert 1973 Afriat and revealed preference theory (Review of Economic Studies 40); Varian 1982 The nonparametric approach to demand analysis (Econometrica 50)

References

Samuelson — A note on the pure theory of consumer's behaviour · Economica 5, 61-71 (1938); originator of the revealed-preference approach
Afriat — The construction of utility functions from expenditure data · Quarterly Journal of Economics 80, 643-664 (1967); originator of Afriat's theorem and the Afriat inequalities
Diewert — Afriat and revealed preference theory · Review of Economic Studies 40, 419-425 (1973); the modern linear-programming proof of Afriat's theorem
Varian — The nonparametric approach to demand analysis · Econometrica 50, 945-973 (1982); originator of the GARP test and the nonparametric revealed-preference programme
Mas-Colell, Whinston, and Green — Microeconomic Theory · Oxford 1995, Ch. 2 (consumer theory), Ch. 3 (aggregate demand), and the mathematical appendix on the integrability problem
Houthakker — Revealed preference and the utility function · Economica 17, 159-174 (1950); originator of the Strong Axiom of Revealed Preference (SARP)

Estimated time

beginner: 15m
intermediate: 40m
master: 80m