Sonnenschein–Mantel–Debreu Theorem

Sonnenschein–Mantel–Debreu theorem (often abbreviated as the SMD theorem or the anything-goes theorem) is a foundational characterization result in mathematical general equilibrium theory and business and economics theory established through sequential proofs by Hugo Sonnenschein (1972, 1973), Rolf Mantel (1974), and Gérard Debreu (1974). Operating within the canonical Arrow–Debreu model of a pure exchange economy, the theorem demonstrates that even when every individual consumer possesses well-behaved, continuous, monotone, and convex preferences that satisfy the individual Slutsky equation and the Weak Axiom of Revealed Preference (WARP), aggregating individual excess demands across heterogeneous consumers with positive endowments preserves only three structural properties at the market level: continuity, homogeneity of degree zero in prices, and Walras's law. Conversely, given any compact subset of positive normalized prices and any continuous vector field on the price sphere that satisfies Walras's law, there exists a populated exchange economy of standard utility-maximizing agents whose aggregate excess demand function coincides with that function on the compact set.

Individual Excess Demand and the Slutsky Decomposition

Consider a pure exchange economy with L ≥ 2 commodities indexed by ℓ ∈ {1, …, L} traded at a positive price vector p = (p₁, …, p_L) ∈ ℝ₊₊^L, and I consumers indexed by i ∈ {1, …, I}. Each consumer i is endowed with an initial commodity bundle ωⁱ ∈ ℝ₊₊^L and maximizes a continuous, monotone, quasi-concave utility function uⁱ(xⁱ) subject to the Walrasian budget constraint p · xⁱ ≤ p · ωⁱ. Subtracting initial endowments from Marshallian demand xⁱ(p, p · ωⁱ) yields consumer i's individual excess demand function zⁱ: ℝ₊₊^L → ℝ^L:

At the individual level, utility maximization imposes stringent mathematical structure on zⁱ(p). Differentiating zⁱ(p) with respect to prices yields the individual Slutsky equation Jacobian matrix Dzⁱ(p):

where Sⁱ(p) is the symmetric, negative semi-definite Hicksian substitution matrix (satisfying v^⊤Sⁱ(p)v ≤ 0 for all v ∈ ℝ^L and Sⁱ(p)p = 0), and the second term is the endowment income effect, which is proportional to net trade z_kⁱ(p). Because of this structure, every individual consumer's excess demand obeys the Weak Axiom of Revealed Preference (if p · zⁱ(q) ≤ 0 and zⁱ(p) ≠ zⁱ(q), then q · zⁱ(p) > 0).

Failure of Structural Inheritance in Aggregate Excess Demand

Summing across all I consumers defines the economy's aggregate excess demand function Z: ℝ₊₊^L → ℝ^L:

A Walrasian general equilibrium is a price vector p^* ∈ ℝ₊₊^L that clears all markets simultaneously: Z(p^*) = 0. Regardless of preference heterogeneity or endowment distribution, linear summation preserves three elementary axioms from individual budget sets:

1. Continuity: Z(p) is a continuous mapping on ℝ₊₊^L.
2. Homogeneity of Degree Zero: Scaling all nominal prices by any scalar λ > 0 leaves every budget constraint unaltered, so Z(λp) = Z(p) for all λ > 0.
3. Walras's Law: Because every consumer exhausts their budget constraint (p · xⁱ(p) = p · ωⁱ) under monotonicity, the aggregate market value of excess demand is identically zero at every price vector:

Summing the individual Slutsky equations across all I consumers reveals why no additional restrictions survive aggregation:

While the aggregate substitution matrix S(p) = ∑_i Sⁱ(p) remains symmetric and negative semi-definite, the aggregate income-effect matrix A(p) = ∑_i (∇_wxⁱ)(zⁱ)^⊤ is a sum of rank-one outer products across buyers (z_kⁱ > 0) and sellers (z_kⁱ < 0) with heterogeneous marginal propensities to consume ∇_wxⁱ. Even at an equilibrium price vector p^* where aggregate net trade vanishes (∑_i zⁱ(p^*) = 0), individual net trades zⁱ(p^*) are nonzero whenever agents trade with one another, allowing the aggregate income-effect matrix A(p^*) to be asymmetric and positive definite in arbitrary directions.

Because Z(p) is homogeneous of degree zero, prices can be normalized onto the positive unit sphere S₊₊^L−1 = { p ∈ ℝ₊₊^L : ‖p‖₂ = 1 }. On S₊₊^L−1, Walras's law p · Z(p) = 0 has a direct differential-geometric interpretation: the vector Z(p) is orthogonal to the position vector p at every point, meaning that aggregate excess demand is a continuous tangent vector field on the positive unit sphere.

• Sonnenschein (1972, 1973) proved that arbitrary polynomials satisfying Walras's law can be locally generated as aggregate excess demands.
• Mantel (1974) proved that any twice continuously differentiable (C²) function satisfying homogeneity and Walras's law on a compact price domain can be generated by 2L consumers with homothetic preferences (strengthened in 1976 to L homothetic consumers with independent endowments).
• Debreu (1974) delivered the sharp geometric proof in Journal of Mathematical Economics, lowering the required number of consumers to at most L agents (matching the number of commodities) and requiring only continuity (C⁰) of the target function.

Theorem (Debreu, 1974): Let f: S₊₊^L−1 → ℝ^L be any continuous function satisfying homogeneity and Walras's law (p · f(p) = 0 for all p ∈ S₊₊^L−1), and let K_ε = { p ∈ S₊₊^L−1 : p_ℓ ≥ ε for all ℓ = 1, …, L } be any compact price subset bounded away from zero by ε > 0. Then there exist I = L consumers, each possessing a continuous, monotone, quasi-concave utility function uⁱ: ℝ₊^L → ℝ and a positive endowment vector ωⁱ ∈ ℝ₊₊^L, whose individual excess demands zⁱ(p) sum to f(p) identically on K_ε:

Debreu's constructive proof decomposes any continuous tangent vector field f(p) on K_ε into a sum of L elementary rank-one excess demand fields gⁱ(p) = θ_i(p)[ eⁱ − (p_i / ‖p‖²)p ], where eⁱ is the i-th canonical basis vector and θ_i(p) > 0 is a continuous scalar weight function, and invokes revealed preference integrability to construct a concave utility function uⁱ generating each gⁱ(p) on K_ε.

Multiplicity of Equilibria and Failure of Walrasian Tâtonnement

Although the Arrow–Debreu model establishes the existence of at least one Walrasian equilibrium Z(p^*) = 0 via Brouwer's or Kakutani's fixed-point theorem, the Sonnenschein–Mantel–Debreu theorem proves that microeconomic rationality alone ensures neither uniqueness nor dynamic stability of equilibrium:

1. Arbitrary Multiplicity of Equilibria: Andreu Mas-Colell (1977) proved that for any non-empty compact subset E ⊂ K_ε, there exists a well-behaved exchange economy whose set of Walrasian equilibrium prices is E (and if E consists of regular equilibria with non-singular Jacobians, Gerard Debreu's 1970 Regular Economies Theorem plus the Poincaré–Hopf index theorem requires only that the number of equilibria be finite and odd).
2. Global Instability of Walrasian Tâtonnement: Léon Walras proposed that an auctioneer adjusts prices proportionally to excess demand via dp_t / dt = Z(p_t). Herbert Scarf (1960) constructed a 3-agent, 3-commodity exchange economy in which the unique Walrasian equilibrium is globally unstable and prices orbit on closed limit cycles; by the SMD theorem, for L ≥ 4 commodities, tâtonnement dynamics can embed strange attractors and deterministic chaos.

Critique of the Representative Agent and Escape Routes

In macroeconomics and the Ramsey–Cass–Koopmans model, researchers frequently model the household sector as a single utility-maximizing representative agent. Alan Kirman (1992) emphasized that by Terence Gorman's (1953) aggregation theorem, an economy of heterogeneous consumers behaves as a single representative consumer whose preferences are independent of income distribution if and only if every consumer's indirect utility takes the Gorman polar form vⁱ(p, wⁱ) = αⁱ(p) + β(p)wⁱ, requiring all households to share identical linear Engel curves with the same marginal propensity to consume ∂x_ℓⁱ / ∂wⁱ = b_ℓ(p).

To obtain unique, stable equilibria without a Gorman representative agent, economists impose structural conditions such as Gross Substitutability (Arrow, Block, and Hurwicz, 1959: ∂Z_ℓ / ∂p_k > 0 for all ℓ ≠ k, which ensures WARP in the aggregate and global stability), collinear endowments around Pareto-efficient allocations, or statistical distributional regularity across household expenditure densities (Hildenbrand, 1983, 1994).

Does the Sonnenschein–Mantel–Debreu theorem mean that general equilibrium does not exist?

No. The existence of at least one Walrasian general equilibrium Z(p^*) = 0 is established by the Arrow–Debreu (1954) existence theorem, which relies only on continuity, homogeneity of degree zero, Walras's law (p · Z(p) = 0), and boundary survival conditions as prices approach zero outside the interior compact set K_ε. What the SMD theorem disproves is that individual rationality ensures uniqueness, stability, or Aggregate Revealed Preference inside the interior of the price simplex.

Why don't individual downward-sloping substitution effects add up to a well-behaved market excess demand curve?

By the Slutsky equation, a change in the price p_k of a commodity affects each consumer i through two channels: a negative own-price substitution effect (S_kkⁱ ≤ 0) and an endowment income effect (−(∂x_kⁱ / ∂wⁱ)z_kⁱ). When the price of wheat rises, net sellers of wheat (z_kⁱ < 0) become wealthier while net buyers (z_kⁱ > 0) become poorer. If net sellers have a higher marginal propensity to consume wheat than net buyers, the positive wealth effect on sellers can outweigh both the negative wealth effect on buyers and the aggregate substitution effect, causing total market excess demand for wheat to rise when its price increases.

Why are at least L consumers needed in Debreu's (1974) theorem?

Each consumer i contributes a rank-one income-effect matrix (∇_wxⁱ)(zⁱ)^⊤ to the aggregate Jacobian DZ(p). On the (L − 1)-dimensional normalized price sphere, spanning an arbitrary asymmetric (L − 1) × (L − 1) Jacobian matrix at every point requires at least L independent rank-one directions; if an economy contains fewer than L consumers (I < L), Chiappori and Ekeland (1999) proved that aggregate excess demand satisfies testable partial differential rank restrictions.

1. Sonnenschein, H. (1972). "Market Excess Demand Functions." Econometrica, 40(3), 549–563. https://doi.org/10.2307/1913184
2. Mantel, R. R. (1974). "On the Characterization of Aggregate Excess Demand." Journal of Economic Theory, 7(3), 348–353. https://doi.org/10.1016/0022-0531(74)90100-890100-8)
3. Debreu, G. (1974). "Excess Demand Functions." Journal of Mathematical Economics, 1(1), 15–21. https://doi.org/10.1016/0304-4068(74)90032-990032-9)
4. Kirman, A. P. (1992). "Whom or What Does the Representative Individual Represent?" Journal of Economic Perspectives, 6(2), 117–136. https://doi.org/10.1257/jep.6.2.117
5. Shafer, W., & Sonnenschein, H. (1982). "Market Demand and Excess Demand Functions." In K. J. Arrow & M. D. Intriligator (Eds.), Handbook of Mathematical Economics, Vol. 2 (pp. 671–693). North-Holland. https://doi.org/10.1016/S1573-4382(82)02009-802009-8)