General Equilibrium

The Arrow-Debreu framework

A general-equilibrium economy specifies:

• L commodities. Distinguished by physical characteristics, location, date, and (in the Arrow-Debreu extension) state of nature. With state-contingent commodities, a single time-zero spot market for each state-of-the-world covers the entire dynamic problem.
• I consumers. Each with preferences ≿_i, endowment ω_i ∈ ℝ₊^L, and profit shares θ_ij.
• J firms. Each with a production set Y_j ⊂ ℝ^L.

A competitive (Walrasian) equilibrium is a price vector p* ∈ ℝ₊^L and allocation (x*, y*) such that:

1. Each consumer maximizes preferences subject to her budget at prices p*.
2. Each firm maximizes profit on Y_j at p*.
3. All markets clear: Σ x*_i = Σ ω_i + Σ y*_j.

The triplet (p*, x*, y*) is the equilibrium. Everyone is individually rational; markets simultaneously balance; the simultaneity is the substance.

The excess-demand function z(p) collects net market imbalances. z(p) is continuous on the open positive orthant, homogeneous of degree zero, satisfies Walras's Law p · z(p) = 0, and is bounded below (consumers cannot supply more than their endowments). Existence reduces to: does there exist a price vector with z(p*) ≤ 0?

Existence: Brouwer's fixed-point theorem

The cleanest proof goes via the price simplex Δ = {p ∈ ℝ₊^L : Σ p_l = 1}. Define a continuous map T: Δ → Δ that increases the share of any good in excess demand:

T_l(p) = (p_l + max{z_l(p), 0}) / (1 + Σ max{z_k(p), 0})

By Brouwer's fixed-point theorem (any continuous map from a non-empty compact convex set to itself has a fixed point), T has a fixed point p* with T(p*) = p*. At a fixed point, no relative-price share wants to grow — meaning every z_l(p*) is non-positive. Walras's Law p* · z(p*) = 0 then forces z_l(p*) = 0 wherever p_l* > 0. Equilibrium.

The full Arrow-Debreu (1954) proof uses Kakutani's fixed-point theorem because demand correspondences (not single-valued functions) appear when preferences are merely convex rather than strictly convex. The technical machinery includes:

• Compactness via truncation of the consumption set.
• Upper hemi-continuity of demand correspondences, following from Berge's maximum theorem.
• Convex-valued demand correspondences, requiring convex preferences.
• Walras's Law in the inequality form p · z(p) ≤ 0.

These ingredients combined feed Kakutani's theorem, yielding existence.

Historical context

Léon Walras (1874-1877) first conceived of an economy in general equilibrium and counted equations and unknowns, but did not prove existence rigorously. Abraham Wald (1936) supplied the first rigorous existence proof for a specialized model. John von Neumann's growth model (1937) used fixed-point arguments. The 1944 von Neumann-Morgenstern game-theoretic apparatus and Kakutani's 1941 fixed-point theorem set the stage.

The breakthrough came in 1954: Kenneth Arrow and Gerard Debreu jointly published Existence of an Equilibrium for a Competitive Economy in Econometrica. Lionel McKenzie published a contemporaneous proof. Debreu's Theory of Value: An Axiomatic Analysis of Economic Equilibrium (1959) consolidated the framework. The state-contingent commodities trick (Arrow, 1953) extended the static framework to handle uncertainty.

The 1970s brought the Sonnenschein-Mantel-Debreu results (Sonnenschein 1972, Mantel 1974, Debreu 1974): aggregate excess-demand functions can be essentially arbitrary continuous functions satisfying Walras's Law. This was a critique of the framework's predictive power — the model is highly general but produces few testable comparative-static results.

Computable general equilibrium (CGE) models emerged in the 1960s-70s (Johansen, Adelman-Robinson, Shoven-Whalley) to solve Arrow-Debreu-style economies numerically for policy analysis. Modern dynamic-stochastic general equilibrium (DSGE) macroeconomics extends the framework with intertemporal optimization and rational expectations.

Worked example: 2-good, 2-consumer exchange

Two consumers (A, B) with utilities U_A(x, y) = x^2/3 y^1/3 and U_B(x, y) = x^1/3 y^2/3. Endowments: ω_A = (10, 5), ω_B = (5, 10). Set p_y = 1.

Step	Anna (Cobb-Douglas α=2/3)	Ben (Cobb-Douglas α=1/3)
Income	10px + 5	5px + 10
Demand for X	(2/3)(10px + 5)/px	(1/3)(5px + 10)/px
Total X demand	(2/3)(10px+5)/px + (1/3)(5px+10)/px
Total X supply	15
Market-clearing	Simplify: 20px/3 + 10/3 + 5/3 + 10/(3px) = 15
Solve	20px/3 + 10/(3px) = 30/3 → 20px² − 30px + 10 = 0
Equilibrium price	px* = 1 (the other root is 0.5, rejected here)
Equilibrium allocation	x_A = (2/3)(15) = 10; y_A = (1/3)(15) = 5	x_B = (1/3)(15) = 5; y_B = (2/3)(15) = 10

At p_x* = 1 (with p_y = 1), market X clears (10 + 5 = 15 = supply). By Walras's Law, market Y also clears. Equilibrium is the endowment itself in this symmetric case — no trade occurs, but the algorithm still verifies that the endowment is an equilibrium. Skewed endowments would produce non-trivial equilibrium prices and allocations.

General vs partial equilibrium and related approaches

	General equilibrium	Partial equilibrium	DSGE	CGE	Agent-based
Markets analyzed	All simultaneously	One (ceteris paribus)	All (dynamic)	All (numerical)	All (simulated)
Origin	Walras, 1874; Arrow-Debreu, 1954	Marshall, 1890	Kydland-Prescott, 1982	Johansen, 1960; Shoven-Whalley	Schelling, 1971; ABM 1990s
Existence	Fixed-point theorems	Curve intersection	Numerical	Numerical	Simulation
Closed form	Rare	Common	Rare	Rare	No
Policy use	Theoretical benchmark	Single-market analysis	Macroeconomic forecasting	Trade, tax, climate policy	Financial, behavioral
Captures externalities	Modeled explicitly	Limited	Yes	Yes	Naturally

General equilibrium is the theoretical benchmark; CGE and DSGE are its computational descendants; agent-based modeling relaxes its equilibrium assumption entirely.

Assumptions, exactly

• Complete markets. A market exists for every commodity, including state-contingent goods (Arrow securities) under uncertainty.
• Convex preferences and production. Needed for demand correspondences to be well-behaved and for the Second Welfare Theorem.
• Continuity of preferences and production. Needed for fixed-point theorems.
• Locally non-satiated preferences. Forces budget constraints to bind; underlies Walras's Law.
• Price-taking. No market power; consumers and firms treat prices as parameters.
• No externalities. One agent's actions don't directly enter others' utility or production functions.
• Walras's Law. p · z(p) ≤ 0, following from budget constraints.

Common misconceptions

• "General equilibrium describes real markets." It is a logical-mathematical model. Real economies satisfy its assumptions only approximately and only in some sectors. The framework is a benchmark, not a literal description.
• "Existence implies stability." The proof shows an equilibrium exists; it does not show tâtonnement or any other dynamic process converges to one. Stability is a separate (often false) result.
• "Equilibrium is unique." Not without strong further assumptions (gross substitutability, weak axiom of revealed preference at the aggregate level). The model permits multiple equilibria.
• "The framework can't handle time or uncertainty." The Arrow-Debreu extension treats commodities as state- and date-contingent. Conceptually elegant; practically requires complete markets for every state-date pair, which real economies lack.
• "Sonnenschein-Mantel-Debreu refutes GE." It shows aggregate excess demand has few restrictions beyond Walras's Law, limiting predictive power — but doesn't refute existence or welfare results. It tells us GE provides fewer comparative-static predictions than partial equilibrium suggests.
• "The Walrasian auctioneer is realistic." No. It is a thought experiment, a heuristic for thinking about how equilibrium prices might be reached. Modern microstructure replaces the auctioneer with order books, market makers, and learning processes.

Applications

• Welfare theorems. Both fundamental welfare theorems are proved within the Arrow-Debreu framework; their interpretation of efficiency rests on its assumptions.
• Computable general equilibrium (CGE). Numerical solution of Arrow-Debreu economies for trade, tax, and climate policy analysis. Models like GTAP, MIRAGE, EPPA simulate millions of equilibrium prices.
• DSGE macroeconomics. Dynamic stochastic general equilibrium models extend Arrow-Debreu to intertemporal optimization with rational expectations. Standard tool at central banks and finance ministries.
• International trade theory. Heckscher-Ohlin, factor-price equalization, and gains-from-trade results all derive within general-equilibrium models. Modern quantitative trade theory (Eaton-Kortum) is GE-style.
• Climate economics. Integrated assessment models (DICE, RICE) embed climate dynamics into Arrow-Debreu-style economies to evaluate carbon-pricing policies.
• Finance. Asset-pricing models (Lucas tree, Merton ICAPM) are general-equilibrium constructions. The risk-neutral measure is the dual of equilibrium state prices.
• Mechanism design. The Arrow-Debreu framework is the benchmark against which alternative institutions are evaluated for welfare properties.