RESEARCH DOSSIER

Existence Stability and Adjustment of Economic Equilibria

Abstract. The mathematical existence of competitive equilibrium, determinacy of relative prices, stability under a designated adjustment process, and empirical attainment of market-clearing allocations constitute distinct research questions. This monograph examines their connections within finite-dimensional competitive exchange and production economies. It reconstructs the logical role of the Arrow–Debreu existence program, the regularity of equilibria, the Sonnenschein–Mantel–Debreu restrictions on aggregate excess demand, and Scarf's demonstrations of nonconvergent price motion. Explicit existence arguments, Jacobian-based local analysis, autonomous counterexamples, discrete-time calculations, and an elementary budget-preserving exchange example establish the difference between consistency and attainability. Experimental studies and an institutional example from organized electricity markets show how different trading rules and information structures affect the interpretation of price adjustment. The original contribution is an integrated methodological synthesis of existence, selection, dynamics, implementation, empirical identification, and welfare assessment. It distinguishes analytical illustrations from established economic realizability results and treats institutional mechanisms as part of the object of research. The study provides a foundation for the interpretation and verification of equilibrium claims without asserting a new theorem, empirical estimate, or named theoretical system.

Keywords: General equilibrium; competitive equilibrium; excess demand; existence; local uniqueness; regular economies; Walras' law; tâtonnement; gross substitutes; dynamic stability; equilibrium selection; market design; computational economics; disequilibrium

Introduction

Economic equilibrium is a condition of joint compatibility among individual plans, resource constraints, technologies, prices, and the institutional rules of exchange. In a competitive economy, buyers and sellers make choices taking the prevailing price system as given. A competitive equilibrium associates these choices with a feasible allocation and a price system at which demands and supplies are compatible. The existence question asks whether such an association can be formed for the specified economy. The determinacy question asks whether the set of solutions consists of one relative price system, several isolated systems, or a continuum. The dynamic question asks how a process operating in historical time behaves around or between solutions.

These questions matter because equilibrium reasoning is used for different purposes. A welfare economist may analyze a feasible allocation and its efficient properties. A theorist may study comparative statics across parameter values. A computational economist may solve market-clearing equations for a counterfactual tax. A market designer may construct a rule that produces mutually consistent orders within a particular trading interval. Each activity has its own object of validation. A numerical solution can meet every imposed equilibrium condition while failing to describe an observed trajectory; equally, a stable observed price can coexist with rationing or incomplete participation.

The subject of the present Research Dossier is the mathematical and institutional relationship among these objects. Its main domain is finite-dimensional competitive equilibrium, with additional attention to incomplete markets and changing environments where they illuminate the limits of the basic case. General equilibrium as a broad discipline, fixed-point algorithms as a mathematical specialty, and distinct theories of strategic equilibrium have their own scientific scope. This study integrates the results needed to judge the existence, uniqueness, adjustment, and attainment of competitive equilibrium.

The method combines selected primary-source texts and scholarly abstracts, direct mathematical deduction, counterexamples, and institutional comparison. The literature is purposively selected for its relevance to foundational questions rather than sampled as a systematic meta-analysis. Calculations in the text are illustrative derivations. They supply transparent logical demonstrations and neither estimate numerical characteristics of a national economy nor reconstruct the data of laboratory studies.

Historical development of the equilibrium problem

The Walrasian tradition represented competitive prices as a simultaneous system. A change in the price of one commodity could alter household wealth and expenditure, production profitability, and the demands for other commodities. This interconnectedness generated an analytical problem beyond the intersection of an isolated demand curve and an isolated supply curve. The market-clearing equations had to be compatible with nonnegative commodity quantities, budget constraints, and technological feasibility.

The existence program brought this question into modern mathematical economics. Arrow and Debreu's (1954) existence of competitive equilibrium integrated household and firm choices using a generalized game framework, with admissibility conditions on consumption, production, and resources. Two existence theorems are stated, including alternative conditions concerning endowments and productive labor. The opening discussion explicitly separates the existence result from uniqueness and the specification of market dynamics.

The stability program examined additional assumptions concerning the movement of prices. The research of Arrow, Block, and Hurwicz (1959) established important conditional results for competitive price adjustment. Scarf's subsequent investigation situated positive convergence results within a narrower class of economies and demonstrated the possibility of oscillation under excess-demand adjustment. The importance of Scarf's instability examples (1960) is that unstable motion can arise from individually optimizing agents, rather than from a rejection of the maximizing assumptions.

A related development concerned the number and structure of competitive equilibria. Debreu (1970) investigated finite equilibrium sets in a differentiable pure-exchange setting. The analysis connects equilibrium prices to regular values of a map defined over the space of economies. In parallel, the aggregate demand literature established the substantial flexibility of market excess demand even when individual demand satisfies conventional rationality conditions. These findings changed the kinds of restrictions necessary to derive robust price predictions.

The subsequent history encompasses numerical fixed-point procedures, path-following price adjustment, experimental economics, and market design. Each branch addresses a different gap in the original simultaneous system. The resulting body of research is unified by the question of which mathematical and institutional conditions suffice for a particular equilibrium claim. It is therefore misleading to read the history as a single sequence in which an existence theorem was later superseded by an instability theorem. The former establishes nonemptiness under stated conditions; the latter examines the behavior of a specified process.

Competitive economies and the meaning of existence

Primitives and individual choices

Let there be nn commodities, indexed by jj, mm households, indexed by ii, and kk firms, indexed by ff. Commodities may distinguish location, delivery date, and contractual state when these distinctions correspond to available markets. Household ii has a consumption set XiX_i, preferences represented where appropriate by uiu_i, an endowment ωi\omega_i, and shares θif\theta_{if} in firms. Firm ff has a production set YfY_f whose members are net output vectors, including negative inputs. A price vector pp specifies prices in a common unit.

A household's competitive choice solves

max⁡xi∈Xiui(xi)subject top⋅xi≤p⋅ωi+∑fθifπf(p).\max_{x_i\in X_i} u_i(x_i) \quad\text{subject to}\quad p\cdot x_i\le p\cdot\omega_i+\sum_f\theta_{if}\pi_f(p).

Here πf(p)=sup⁡y∈Yfp⋅y\pi_f(p)=\sup_{y\in Y_f}p\cdot y, with a finite attained maximum at the equilibrium price under the model's admissibility conditions. The shares for each firm sum to one. Firms choose yfy_f that attain their respective profit maxima. This arrangement incorporates the distribution of profits into household budgets.

Feasibility with free disposal is

∑ixi≤∑iωi+∑fyf.\sum_i x_i\le \sum_i\omega_i+\sum_f y_f.

A competitive equilibrium consists of individual optimal choices and a nonzero nonnegative supporting price vector satisfying feasibility and complementary slackness for surplus commodities. At strictly positive prices, the feasibility inequalities are equalities. A zero price can accompany unused abundance, but the corresponding consumer problem must still have a well-defined optimum. The treatment of zero prices is consequently part of the existence problem, rather than an incidental convention.

Relative prices and market clearing

Multiplying every price by a positive number leaves competitive budget sets unchanged when incomes and profits scale with prices. Hence the relevant price object is usually a ray, represented either on a simplex or in a numeraire chart. A normalization removes price-level indeterminacy attributable solely to the choice of unit. It does not establish economic uniqueness.

For fixed preferences and endowments, and single-valued demand and supply, define excess demand by

z(p)=∑ixi(p)−∑iωi−∑fyf(p).z(p)=\sum_i x_i(p)-\sum_i\omega_i-\sum_f y_f(p).

Under local nonsatiation that entails budget exhaustion, finite profits, and the ownership condition, multiplication by prices gives p⋅z(p)=0p\cdot z(p)=0. This is Walras' law. When all prices are positive, clearing of n−1n-1 markets entails clearing of the remaining market. At a zero-priced commodity, the same argument does not force its excess demand to vanish; feasibility must be checked directly.

Some economies have demand or supply correspondences instead of functions. Multiple maximizing choices are then possible at a given price. An existence proof can establish a compatible selection from those correspondences. Replacing a correspondence by an arbitrary discontinuous numerical selection may destroy the continuity needed by a solver. The distinction between single-valued aggregate demand and general equilibrium through correspondences is thus material.

Conditions and boundary behavior

Existence requires assumptions that control choice, feasibility, continuity, and the boundary of the price domain. In particular, unrestricted profit opportunities, budgets that fail to support acceptable consumption, nonconvex production, indivisibilities, and discontinuous preferences can make common existence proofs inapplicable. Their failure by itself does not prove nonexistence in a particular economy. It identifies which proof must be replaced or which structural feature requires independent analysis.

The relevant conditions differ among exchange models, production models, and models with incomplete contingent markets. The first theorem of Arrow and Debreu permits a particular endowment condition, while its alternative theorem uses conditions on productive labor. These are sufficient conditions within their model. Interpreting them as necessary conditions for all market economies would exceed the scope of the article.

A constructive existence argument in pure exchange

A simplified argument clarifies the function of compactness and the price boundary. Let e=∑iωie=\sum_i\omega_i be strictly positive componentwise. Suppose aggregate excess demand zz is single-valued and continuous on the strictly positive price domain, homogeneous of degree zero, satisfies Walras' law, and is bounded below by −e-e. Assume also that as normalized positive prices approach any boundary point of the closed simplex, at least one component of excess demand becomes arbitrarily large along that sequence in the sense that max⁡jzj(p)→+∞\max_j z_j(p)\to+\infty. This boundary condition is stronger than necessary and is expressly adopted for the demonstration.

For 0<ε<1/n0<\varepsilon<1/n, define the compact convex truncated simplex

Δε={p:∑jpj=1, pj≥ε}.\Delta_\varepsilon=\{p:\sum_jp_j=1,\ p_j\ge\varepsilon\}.

Consider the continuous map Tε(p)=PΔε(p+z(p))T_\varepsilon(p)=P_{\Delta_\varepsilon}(p+z(p)), where PP is Euclidean projection. Its domain and range are the same compact convex set. Brouwer's theorem supplies a fixed point pεp^\varepsilon. The projection characterization gives

z(pε)⋅(q−pε)≤0for all q∈Δε.z(p^\varepsilon)\cdot(q-p^\varepsilon)\le0 \quad\text{for all }q\in\Delta_\varepsilon.

Using Walras' law, take qq with its jjth component equal to 1−(n−1)ε1-(n-1)\varepsilon and every other component equal to ε\varepsilon. Then

(1−nε)zj(pε)+ε∑ℓzℓ(pε)≤0.(1-n\varepsilon)z_j(p^\varepsilon) +\varepsilon\sum_\ell z_\ell(p^\varepsilon)\le0.

The lower bounds imply ∑ℓzℓ≥−∑ℓeℓ\sum_\ell z_\ell\ge-\sum_\ell e_\ell, hence

zj(pε)≤ε∑ℓeℓ1−nε.z_j(p^\varepsilon)\le \frac{\varepsilon\sum_\ell e_\ell}{1-n\varepsilon}.

As ε\varepsilon tends to zero, compactness supplies a convergent subsequence. The upper bound rules out a boundary limit by the adopted divergence assumption. The limit p∗p^* is therefore interior and continuity implies zj(p∗)≤0z_j(p^*)\le0 for every jj. Walras' law and strict positivity yield z(p∗)=0z(p^*)=0.

This is an existence proof for the stated reduced exchange economy, not a reproduction of the general Arrow–Debreu theorem. It shows why a fixed point of a price-update map requires an additional argument connecting it to a genuine equilibrium, and why unbounded excess demand at vanishing prices cannot be ignored by simply writing a map on the closed simplex. The projection apparatus is a mathematical device; its iterative application need not converge. Brouwer asserts that the fixed point exists, even when naïve iteration moves away from it.

A useful economic interpretation is that the truncation prevents the proof from evaluating ill-defined demand at zero prices. The limit argument ensures that the artificial price floor does not determine the final solution. An applied model that retains a positive floor permanently solves a related constrained equilibrium problem. The proof makes the difference explicit.

Uniqueness regularity and comparative statics

Local isolation

In an interior numeraire chart write q=(p1/pn,…,pn−1/pn)q=(p_1/p_n,\ldots,p_{n-1}/p_n) and denote the first n−1n-1 components of excess demand by F(q,η)F(q,\eta), where η\eta collects economic parameters. At an equilibrium (q∗,η∗)(q^*,\eta^*), F(q∗,η∗)=0F(q^*,\eta^*)=0. If FF is continuously differentiable and its Jacobian with respect to qq,

J=DqF(q∗,η∗),J=D_qF(q^*,\eta^*),

is invertible, the implicit function theorem yields a unique differentiable equilibrium branch in a neighborhood of the point. Relative-price comparative statics on that branch obey

Dηq∗=−J−1DηF.D_\eta q^*=-J^{-1}D_\eta F.

This derivative is meaningful locally and depends on the selected regular equilibrium. If another branch exists outside the neighborhood, local comparative statics cannot establish which branch an economy will reach after a large change.

The condition of nonsingularity is sufficient for local isolation. Failure of it leaves open whether isolation nevertheless holds. For example, an equation proportional to (q−q∗)3(q-q^*)^3 has an isolated zero with zero derivative. Near singularities a linear approximation may give poor forecasts; second- or higher-order terms can govern the behavior.

Debreu's (1970) finite-equilibrium analysis gives a precise setting in which regularity is prevalent outside a closed null exceptional set in the endowment space. It relies on the assumptions of the specified differentiable exchange environment and a finite-dimensional parameterization. The finiteness of a compact discrete set follows by contradiction: an infinite subset of a compact space has an accumulation point, whereas local isolation rules out such a sequence within the closed equilibrium set.

Multiplicity and sensitivity

Finite multiplicity remains economically consequential. Equilibria may involve different relative prices, allocations, and welfare, and may respond differently to the same shock. Generic finiteness is compatible with an open region of economies possessing multiple equilibria. A statistical distribution over endowments is needed to convert measure-theoretic genericity into a probabilistic statement about an empirical population; even then its support could concentrate near singular sets.

A derivative that is large in magnitude may indicate economic sensitivity or poor numerical conditioning. The distinction depends partly on units. For a residual equation FF in physical commodities, rescaling tons to kilograms rescales its Jacobian. Robust numerical work should therefore report dimensionless scaling, economically calibrated perturbations, and relative condition measures. An exact derivative in a badly conditioned specification may be correct but uninformative at the precision of the data.

Aggregate excess demand and its limits

Individual optimization imposes restrictions on demand. Their preservation under aggregation is a separate mathematical question because wealth and consumption patterns vary across households. In pure exchange, income p⋅ωip\cdot\omega_i is itself affected by prices. Even if compensated individual substitution satisfies the usual restrictions, uncompensated aggregate responses include changes in purchasing power associated with ownership of the commodities.

The research lineage begins with Sonnenschein's characterization question (1973), continued by Mantel's aggregate-demand analysis (1974) and Debreu's excess-demand result (1974). Their contributions differ in regularity, domain, and decomposition techniques. The contemporary community excess demand characterization of McFadden, Mas-Colell, Mantel, and Richter (1974) is especially useful for comparing these results: its discussion distinguishes polynomial approximations, differentiable decompositions, and continuous functions on compact positive-price domains.

The broad lesson is that continuity, homogeneity of degree zero, Walras' law, and appropriate lower bounds or boundary restrictions leave considerable freedom for aggregate excess-demand behavior. On compact sets within the positive orthant, suitable candidate functions can be represented by economies with optimizing consumers. A local construction of an excess-demand field satisfying the elementary identities is consequently plausible as illustrative reduced-form mathematics. To claim a fully realized competitive economy, however, one must invoke a representation theorem with verified domain and assumptions or exhibit the individual preferences and endowments directly.

The theorem family does not say that any arbitrary dynamic is the behavior of every exchange economy. It is a characterization across a class of constructed economies. It also does not eliminate restrictions introduced by common preferences, specific endowment patterns, technologies, observations across multiple wealth distributions, or data on individual decisions. Economists can obtain testable implications by specifying and independently identifying such restrictions. What cannot be inferred from individual rationality alone is a universal aggregate slope or universal convergence of unqualified price adjustment.

The findings expose a problem with treating a definite representative household as automatic. Aggregation into a representative optimization problem requires restrictions on preferences, wealth distribution, and the way individual demands combine. A representative household may be a useful model construction. Its predicted stability should be interpreted as conditional on its restrictions rather than as an unrestricted consequence of many heterogeneous households optimizing.

Price adjustment and local stability

Specify the motion

A price adjustment process is a law mapping a state and its economic environment into a direction of change. In a continuous-time numeraire chart, write

q˙=A(q,η)F(q,η).\dot q=A(q,\eta)F(q,\eta).

The matrix AA converts excess demand measured in commodity units into relative price velocities and records the adjustment mechanism. It may be diagonal with positive entries for own-market responses or include cross-market feedback. A normalized full-price formulation requires an appropriate projection or invariant normalization to ensure that the price path remains in its intended domain. The numeraire formulation here is a chosen rule, not automatically identical to every normalization of full-price tâtonnement.

At an interior equilibrium, if AA and FF are differentiable, the linearization is

h˙=Bh,B=A(q∗,η∗)J.\dot h=B h,\qquad B=A(q^*,\eta^*)J.

The derivative of AA multiplies F(q∗,η∗)=0F(q^*,\eta^*)=0 and so drops out at first order. If all eigenvalues of BB have strictly negative real parts, the equilibrium is locally asymptotically stable. A strictly positive real part establishes local instability. Eigenvalues exactly on the imaginary axis require additional nonlinear analysis, except where the dynamics are exactly linear.

Lyapunov stability means that paths starting sufficiently close remain close for all future time. Attraction means that paths in a neighborhood approach the equilibrium; local asymptotic stability combines these. Global attraction concerns a specified full domain of admissible initial conditions, including its invariant boundaries. Convergence to a set can occur without convergence to one particular equilibrium.

A fully specified exchange example

Consider two households with utilities ui(xi1,xi2)=xi1xi2u_i(x_{i1},x_{i2})=\sqrt{x_{i1}x_{i2}}, endowments (1,0)(1,0) and (0,1)(0,1), respectively, and two goods. At strictly positive prices normalized as (q,1)(q,1), each household spends half its income on each commodity. Their incomes are qq and 11.

Aggregate demand for the first commodity is 1/2+1/(2q)1/2+1/(2q); its available quantity is one. Excess demand is

F(q)=12q−12.F(q)=\frac{1}{2q}-\frac12.

Demand for the second commodity is (q+1)/2(q+1)/2, giving excess demand (q−1)/2(q-1)/2. The two expressions satisfy qF(q)+(q−1)/2=0qF(q)+(q-1)/2=0. They vanish jointly only at q∗=1q^*=1. Each household then consumes (1/2,1/2)(1/2,1/2), and the allocation exhausts both aggregate endowments.

Adopt a multiplicative price rule q˙=κqF(q)\dot q=\kappa qF(q), κ>0\kappa>0. It reduces exactly to q˙=κ(1−q)/2\dot q=\kappa(1-q)/2; the solution is

q(t)=1+(q(0)−1)e−κt/2.q(t)=1+(q(0)-1)e^{-\kappa t/2}.

Any positive initial relative price stays positive and converges to one. The function V(q)=(q−1)2/2V(q)=(q-1)^2/2 has derivative −κ(q−1)2/2-\kappa(q-1)^2/2, strictly negative away from equilibrium. This example proves global attraction in the positive relative-price domain for the specified dynamics and this economy. It illustrates the importance of jointly specifying preferences, endowments, and the update rule.

Changing the speed κ\kappa affects the time required to approach equilibrium while leaving the equilibrium itself unchanged. For this example, half the initial absolute deviation remains after 2log⁡(2)/κ2\log(2)/\kappa units of model time. Such a formula becomes an empirical estimate only after the time unit and speed have been identified from the relevant institution.

Global behavior and counterexamples

Rotational and unstable fields

A simple two-dimensional chart permits an exact comparison among local adjustment mechanisms. Let q∗=(1,1)q^*=(1,1), h=q−q∗h=q-q^*, and consider a linear law h˙=Bh\dot h=B h near h=0h=0. The matrices

Bs=(−100−2),Bu=(1002),Bc=(0−110)B_s=\begin{pmatrix}-1&0\\0&-2\end{pmatrix}, \quad B_u=\begin{pmatrix}1&0\\0&2\end{pmatrix}, \quad B_c=\begin{pmatrix}0&-1\\1&0\end{pmatrix}

are all nonsingular. Each has the unique fixed point h=0h=0. The first has exponentially decaying trajectories; the second has trajectories moving outward; the third preserves ∥h∥2\|h\|^2 because BcB_c is skew-symmetric and produces exact periodic orbits of period 2π2\pi.

These examples can be confined to a neighborhood within the positive-price chart. The third market's excess demand may be defined there by z3=−q1F1−q2F2z_3=-q_1F_1-q_2F_2 with F=BhF=Bh; the resulting field satisfies Walras' law and can be extended homogeneously along positive rays. This is a local reduced-form illustration. No assertion is made that these globally linear fields satisfy every boundary and resource condition for a competitive economy or that the unstable trajectory can continue forever within this chart.

Scarf (1960) supplies the economically stronger result: global instability of competitive equilibrium can be generated by actual optimizing households and a specified excess-demand motion. In his examples, complementarity creates intermarket feedback that an own-market adjustment rule does not damp. The lesson is that a proof of individual maximization and a proof of aggregate convergence involve distinct restrictions.

A sufficient condition in a stated setting

Gross substitutability restricts how demand for one good changes when another good becomes more expensive. For fixed resources in exchange, one common aggregate formulation is ∂zi/∂pj≥0\partial z_i/\partial p_j\ge0 for i≠ji\ne j on the interior, with appropriate strictness or irreducibility for stronger conclusions. Positive convergence results in the research tradition of Arrow, Block, and Hurwicz depend on such conditions and their associated price domain and dynamic assumptions.

One transparent conditional argument can be given for a differentiable pure-exchange field under stronger strict gross substitutability. Suppose all off-diagonal derivatives are strictly positive, the equilibrium p∗p^* is interior, and p˙=z(p)\dot p=z(p) has a globally defined positive solution. Walras' law gives p⋅p˙=0p\cdot\dot p=0, so the Euclidean norm of pp is constant. Scale p∗p^* to that same norm. For pp not proportional to p∗p^*, let M=max⁡i(pi/pi∗)M=\max_i(p_i/p_i^*) and m=min⁡i(pi/pi∗)m=\min_i(p_i/p_i^*).

At a maximizing index ii, pi=Mpi∗p_i=M p_i^*, and the other prices are no greater than the corresponding components of Mp∗M p^*. At least one other price is strictly smaller. By strict gross substitutability, zi(p)<zi(Mp∗)=0z_i(p)<z_i(Mp^*)=0. At a minimizing index the reverse argument gives zi(p)>0z_i(p)>0. Thus MM decreases and mm increases along any non-equilibrium path. The componentwise bounds keep the path inside a compact positive region. Any non-equilibrium limit point would allow a strictly shrinking maximum ratio over a short time, contradicting convergence of the nonincreasing maximum along an invariant limit set. It follows that all prices approach the equilibrium on the same norm sphere.

This sufficient demonstration has explicit stronger assumptions; it is not an assertion that all weak-substitutes economies have unique prices or that every update scheme shares the same stability. The argument shows how order, homogeneity, and an invariant normalization can be combined to prove global behavior. Dropping strictness can require an independent analysis of linked markets and stationary subsets.

Multiple basins and equilibrium selection

For a one-dimensional reduced-form law take h˙=−h(h2−a2)\dot h=-h(h^2-a^2), a>0a>0, in a bounded interval of feasible relative-price deviations. The equilibria are −a,0,a-a,0,a. The derivative (a2−3h2)(a^2-3h^2) equals a2a^2 at zero and −2a2-2a^2 at the two outer points. The sign pattern implies that initial conditions with h>0h>0 within an invariant neighborhood approach aa, whereas those with h<0h<0 approach −a-a. The initial value zero remains there.

This constructed field exhibits selection through basins of attraction. It also separates global convergence to some equilibrium from global attraction to a given equilibrium. The middle solution is isolated and regular yet dynamically unstable. The outer solutions are isolated and locally stable, and their selection depends on history. In economic applications, empirical predictions about selection must therefore identify initial states, shocks, and the updating institution, rather than simply count stationary solutions.

Discrete adjustment delays and changing fundamentals

Discretization is a mechanism choice

Many observed markets update in discrete intervals. Consider the earlier Cobb–Douglas example implemented by the explicit iteration

qt+1=qt+τκqtF(qt)=1+(1−τκ/2)(qt−1).q_{t+1}=q_t+\tau\kappa q_t F(q_t) =1+(1-\tau\kappa/2)(q_t-1).

Its multiplier is r=1−τκ/2r=1-\tau\kappa/2. Local convergence requires ∣r∣<1|r|<1, or 0<τκ<40<\tau\kappa<4. For 0<τκ≤20<\tau\kappa\le2, all positive initial prices remain positive and the deviations decay without sign reversals, except the one-step case at equality where convergence is immediate. For 2<τκ<42<\tau\kappa<4, small enough deviations alternate and decay; unrestricted positive starting prices can produce inadmissible nonpositive iterates and require a separate positivity-preserving rule.

At τκ=4\tau\kappa=4, r=−1r=-1 and small nonzero deviations oscillate without decay. Beyond this threshold, the fixed point is locally repelling, although the continuous-time dynamics remains globally attractive. An unstable numerical update can arise from the size of the update interval rather than a change in economic fundamentals.

The following analytic calculations use initial q0=1.2q_0=1.2 and unrounded multipliers. They are an illustrative table, not observed price data.

τκ\tau\kappa Multiplier q1q_1 q2q_2 Local behavior
1 0.5 1.10 1.05 Monotone convergence
2 0 1.00 1.00 Exact one-step convergence
3 -0.5 0.90 1.05 Damped alternation
4 -1 0.80 1.20 Persistent alternation
5 -1.5 0.70 1.45 Expanding deviation

A related issue arises with response delays. Let h˙(t)=−αh(t−d)\dot h(t)=-\alpha h(t-d), α>0\alpha>0, with a nonnegative delay dd and a prescribed initial history on [−d,0][-d,0]. Substitution of eλte^{\lambda t} gives the characteristic equation λ+αe−λd=0\lambda+\alpha e^{-\lambda d}=0. A purely imaginary root λ=iω\lambda=i\omega requires cos⁡(ωd)=0\cos(\omega d)=0 and ω=αsin⁡(ωd)\omega=\alpha\sin(\omega d). The first positive frequency crossing therefore occurs at αd=π/2\alpha d=\pi/2. The calculation establishes the location of this possible stability boundary; a complete delay-system stability proof also requires control of the other characteristic roots and crossing direction.

Delayed feedback appears when prices respond to orders submitted earlier, when firms need time to expand capacity, or when forecasts are revised periodically. The theoretical delay has a measurable counterpart only when the information timestamps and decision timing match the model. An observed oscillation could equally arise from periodic external shocks.

Moving equilibrium and tracking

Let an equilibrium relative price q∗(t)q^*(t) shift with changing fundamentals. In a simple local linear tracking law q˙=−a(q−q∗(t))\dot q=-a(q-q^*(t)), a>0a>0, the tracking error e(t)=q(t)−q∗(t)e(t)=q(t)-q^*(t) satisfies e˙=−ae−q˙∗\dot e=-ae-\dot q^*. If ∣q˙∗∣≤v|\dot q^*|\le v, integration yields

∣e(t)∣≤∣e(0)∣e−at+va(1−e−at).|e(t)|\le |e(0)|e^{-at}+\frac{v}{a}(1-e^{-at}).

Although each frozen-fundamentals equilibrium is attractive, persistent adjustment error can remain when the target moves. A larger response coefficient improves tracking within this continuous model, while its discrete implementation may overshoot at a large interval. The bound thus links stability, environmental change, and timing without conflating them. In empirical work it warns against interpreting persistent excess demand as proof of an unstable fixed-fundamentals system.

Trade outside equilibrium and institutional design

Walrasian tâtonnement is commonly modeled as price revision prior to definitive trade. Excess demand is computed from plans at provisional prices and endowments remain fixed during the adjustment. This specification separates the problem of finding a clearing price from the distributional consequences of transactions along the path.

Actual trade at provisional prices changes holdings. If the vector of holdings is ω(t)\omega(t) and transaction quantity for household ii is vi(t)v_i(t), a minimal accounting condition is ω˙i=vi\dot\omega_i=v_i and ∑ivi=0\sum_i v_i=0 in pure exchange. Actual transactions must also meet budget, inventory, short-sale, and settlement constraints as relevant. Demand now depends on the current allocation of resource ownership. The price dynamics and the endowment dynamics form a coupled system.

Its equilibrium set may differ from the equilibrium set computed using the original endowments. Even when final allocations are feasible and competitive at final holdings, the distribution of purchasing power can reflect earlier transactions. Path-dependent wealth changes create an explanation for why stability of transaction prices, final allocative efficiency, and distribution of trading gains require separate evidence.

Institutions affect the information and actions available along a path. A simultaneous call auction aggregates orders before execution and typically chooses a clearing rule. Continuous double auctions process bids and offers sequentially, creating transaction-specific prices and updating residual inventories. Bilateral bargaining involves decentralized information and counterparty selection. Administered or regulated prices may incorporate explicit balancing quantities, queues, or settlement provisions. These institutional descriptions specify feasible observables and decision timing; they do not supply a universal ordering of convergence speeds.

The study by Goeree and Lindsay (2016), market design and equilibrium stability, explicitly studies the connection between institutional design and dynamics in a Scarf environment. Its reported replication of unstable double-auction behavior and associated efficiency and distributional effects indicates the relevance of market architecture to empirical interpretation. The available abstract supports the qualitative contrast. Detailed estimates and treatment-level mechanisms would require examination of the full article and data.

An adjustment rule also differs in informational requirements. An own-market excess-demand rule can use limited aggregate signals. A Newton-type step generally requires a derivative matrix or an estimate of how multiple markets jointly respond. A path-following mechanism may use additional state information or specially defined constraints. Mathematical convergence under richer information can coexist with institutional difficulty in collecting it or providing incentives for truthful reports.

Equilibrium computation and verification

Computing a competitive equilibrium means finding an allocation and prices consistent with optimization and feasibility to a specified numerical tolerance. The equilibrium computation program represented by Scarf's (1973) computation of economic equilibria develops approximation procedures for fixed points, Walrasian applications, degeneracy, and simplicial subdivision. Its importance lies in transforming the mathematical existence problem into a constructive numerical problem. The step sequence of a numerical routine is a computational artifact unless the routine is also justified as an actual market adjustment protocol.

Several methods illustrate the distinction. Newton's method for F(q)=0F(q)=0 applies the iteration

qr+1=qr−[DqF(qr)]−1F(qr),q_{r+1}=q_r-[D_qF(q_r)]^{-1}F(q_r),

when the inverse exists. Near a regular sufficiently smooth solution, local convergence can be fast. Away from it the iteration may become undefined, leave the positive-price domain, or converge to a different root. Damping, line searches, or trust regions can improve behavior but require their own safeguards. An approximate derivative from finite differences may introduce serious error near poorly conditioned solutions.

Continuation methods embed an easily solved problem in a parameterized family and follow a solution path. They can handle situations where direct iteration fails. A path can fold in the original parameter, making ordinary continuation inadequate; an arclength formulation may permit traversal of a smooth fold. The procedure still needs a starting solution, path orientation, an adequate regularity profile, and a stopping criterion.

For equilibrium with free disposal, inequalities and complementarity provide a natural representation. Under finite prices and feasible choices, conditions include zj≤0z_j\le0, pj≥0p_j\ge0, and pjzj=0p_jz_j=0, together with a price normalization. Computational success should be measured against these conditions rather than just the norm of a reduced system. At strictly positive prices, the complete vector of market residuals is useful as a diagnostic even though one equation is redundant.

A practical verification report should identify the price normalization, constituent optimality residuals, primal feasibility, budget feasibility, complementarity, aggregation, units and tolerances. For a computed allocation x^i\hat x_i, one can inspect positive budget violations max⁡(0,p⋅x^i−wi)\max(0,p\cdot\hat x_i-w_i) and optimize independently or use suitable first-order or dual certificates subject to convexity assumptions. For market clearing with capacities, the residual should be reported both in physical units and scaled by an economically meaningful positive quantity. Feasibility checks must retain zero-demand goods and binding limits instead of dividing by a zero benchmark.

An equation solver returning a small residual provides a valuable consistency check. It does not automatically establish uniqueness, the absence of missed equilibria, or an economically accurate counterfactual. Multi-start testing can discover additional solutions, though finitely many starts cannot generally certify their absence. An interval proof, analytical monotonicity restriction, or a suitable global certificate supplies stronger evidence when available.

Herings (2024), in globally convergent adjustment processes, surveys several price-path constructions with different initial-price, boundary, and regularity conditions. The convergence claims in that review illustrate why the qualifiers distinguish processes: almost every initial price system, a generic class of economies, and all initial prices within a designated domain refer to different quantifiers. Differential-topological path structure may support a stronger global statement than ordinary local tâtonnement, but the implemented process and its assumptions must be matched to the theorem.

Experimental evidence and causal identification

Experimental economics permits direct manipulation of market institutions, incentives, allocation of initial resources, and sometimes the topology of complementary commodity markets. Induced utilities and endowments offer a known benchmark against which the observed trajectory can be compared. The controlled environment is especially valuable when a nonconvergent price process is difficult to distinguish from a continually shifting equilibrium in ordinary field data.

Anderson, Plott, Shimomura, and Granat's (2004) experimental Scarf economy is identified in the Caltech publication record as a study of general equilibrium instability. The later replication reported by Goeree and Lindsay (2016) supports the qualitative finding of cyclical behavior in double-auction implementations of such an environment. This comparison is literature-based: no transaction-level records have been downloaded or independently analyzed in this monograph.

The abstract of Crockett, Oprea, and Plott's (2011) Gale example in the lab reports support for dynamics in which different initial prices lead toward extreme equilibria with asymmetric distribution of gains. The study makes equilibrium selection empirically salient. When the initial price determines the basin of attraction, initial price setting becomes a distributive feature of the market even if later trading follows a regular mechanism.

Gillen, Hirota, Hsu, Plott, and Rogers (2021) investigate convergence-challenging experimental markets. Their reported structural analysis connects local price changes to excess demand and identifies cross-market effects that vary with the severity of disequilibrium. Their results motivate separate tests of short-horizon directional prediction and long-run convergence: a model may successfully predict the sign of a near-term price movement even when the trajectory fails to approach the stationary point.

A research design for adjustment

Suppose market-level observations include price vectors, submitted order quantities, accepted transactions, inventory, time stamps, and public information. A local empirical specification might relate subsequent price changes to current excess demand after conditioning on an identified normalization and contemporaneous shocks. A researcher should measure excess demand from intended net purchases at a relevant price, not assume that completed purchases minus completed sales measure excess demand. The latter is identically zero within balanced bilateral transactions.

The estimated relation may be endogenous because prospective price changes affect submissions, and submitted orders can be strategic. A controlled shift of valuation schedules, exogenous assignment of endowments, or randomized trading rules can supply useful identifying variation. With field data, the exclusion restrictions must be defended on institutional grounds.

Stability claims require more than a negative average own-price coefficient. The local system's eigenvalues depend on the full cross-market derivative matrix and the adjustment rule. Estimating these from a small sample can make classification highly uncertain, especially near the imaginary axis or unit circle. Sampling uncertainty should propagate to the stability criterion rather than be hidden by a binary label.

Testing convergence from many starts is distinct from evaluating a single observed path. A pre-specified neighborhood and time horizon, sufficient repeated independent market sessions, and evidence on attrition or interventions are important. The finite time of a laboratory session cannot prove the asymptotic behavior of an infinite-horizon process. Its evidential value is that particular predictions about directional changes, fluctuations, and attainment can be rejected or sustained within observed conditions.

Institutional applications and international comparability

Organized electricity markets

An electricity market illustrates the distinction between a clearing outcome and decentralized convergence. Generation and consumption must be coordinated within operational time windows, while geographic networks, capacity constraints, forecast revisions, and losses affect feasible exchange. The Federal Energy Regulatory Commission's energy markets explanation identifies linked day-ahead and real-time markets in organized United States regional systems, with financial commitments and subsequent balancing and dispatch. Locational marginal prices reflect marginal energy cost, losses, and congestion.

Within a stylized convex dispatch problem, minimizing cost subject to nodal balance and network limits produces multipliers on the balance constraints. Under suitable differentiability and regularity, those multipliers indicate the marginal value of an additional unit of energy at the respective nodes. Capacity and transmission constraints explain why identical physical energy can carry different locational values. Actual electricity markets may also contain nonconvex commitment decisions, reliability interventions, uplift payments, and pricing rules; the simplest competitive convex model does not capture all settlements.

The resulting price can be the output of a coordinated optimization within a prescribed interval. Repeated differences between day-ahead and real-time prices can arise because forecasts and feasible generation change. They need not indicate instability around one unchanging equilibrium. Classification requires the relevant time, contractual obligations, network state, and dispatch rules.

Comparative institutional analysis

The general methodological point extends to commodity trading, interbank funding, securities auctions, and markets with import restrictions. International comparison must align commodity definitions, currencies, prices inclusive or exclusive of taxes, delivery times, contractual terms, and the nature of actual trade. Across jurisdictions, a seemingly identical commodity can have different economically relevant specifications because of settlement, transport, or legal constraints.

Different markets use different ways of aggregating orders and enforcing feasibility. A multinational study of adjustment should record whether the observed prices are executed transactions, auction-clearing prices, indicative quotes, administratively specified prices, or contract settlement values. Comparison of unconditional price volatility across these regimes conflates differences in measurement and coordination.

A further problem is that external shocks may be jointly transmitted across markets. The covariance of prices may reflect common energy costs, exchange-rate changes, or changes in financing availability. It does not by itself identify an endogenous excess-demand feedback. Structural identification requires variation that distinguishes shifts of fundamentals from movement generated by the adjustment law.

The present study does not estimate cross-national stability coefficients. Its institutional example is grounded in an official United States source and establishes the concrete importance of market design. General applicability to other jurisdictions is a research proposition whose validation requires their own rules, time-series definitions, and data. The scientific value of comparability arises through this explicit account of what needs to be held constant.

Welfare distribution and policy interpretation

Competitive market clearing can support particular welfare results under assumptions such as local nonsatiation, well-defined individual optimization, and absence of relevant unpriced external effects. In the pure-exchange setting, if each household chooses an optimum in its budget set and a feasible allocation makes everyone weakly better off and someone strictly better off, local nonsatiation and optimality imply that every weakly preferred bundle must cost at least its equilibrium income and a strictly preferred bundle must cost strictly more. Summing these inequalities contradicts aggregate feasibility. This is a standard supporting-price argument for Pareto efficiency.

The argument applies to an equilibrium allocation and says nothing about how a price process reaches it. It also does not imply a particular distribution of welfare. Distinct distributions of endowments can support different efficient allocations; distributional evaluation requires additional criteria such as social welfare weights, rights, or policy constraints stated separately from Pareto efficiency.

With disequilibrium exchange, different paths can change the distribution of holdings and therefore subsequent opportunities. Even when prices later clear markets, the cumulative gains and losses may depend on adjustment history. The evidence from laboratory markets described by Crockett, Oprea, and Plott makes this distinction operational. In an ordinary economy the issue may be amplified by debt contracts, illiquidity, or capacity changes, but the direction and magnitude require institution-specific evidence.

Comparative statics offers a conditional description of equilibria after an exogenous parameter change. It does not provide transitional costs or describe the timing of policy effects unless the model explicitly includes the transition. A tax that raises the welfare of a modeled final allocation could entail substantial adjustment costs under a different institutional process. Conversely, a temporary disturbance cannot establish that the post-policy equilibrium is inferior. Researchers should therefore distinguish allocation effects at a defined endpoint from transition effects measured across a defined interval.

A relevant policy analysis also distinguishes a mechanism intended to stabilize prices from a rule intended to guarantee feasibility. Price ceilings, subsidies, circuit breakers, reserves, and auction settlement rules act through different channels. A policy may lower one measure of volatility while altering quantities, inventories, participation, or the information available for price discovery. The eligible outcomes and welfare criterion need to be specified before any general assessment is defensible.

Analytical synthesis and open problems

The preceding analysis permits a unified interpretation of an equilibrium claim as a sequence of distinct evidentiary tasks. The account begins with the economic primitives that define possible allocations and feasible exchanges. It then describes the relation of relative prices to individual choices and identifies the mathematical conditions that guarantee existence. For a given solution, it examines regularity and multiplicity. It specifies a process in historical or computational time, and only then evaluates local stability, global behavior, and welfare along relevant trajectories.

These stages are logically connected but have different verification methods. The following comparison summarizes their respective evidence.

Research question Adequate type of evidence Further issue requiring independent work
Does a competitive equilibrium exist? A verified existence theorem or explicit solution with admissible primitives Whether it is unique or attainable
Is an equilibrium locally regular? A nonsingular reduced market-clearing Jacobian at an interior solution Global uniqueness and dynamic stability
Does a specified process converge locally? A valid local dynamic analysis with strict spectral or Lyapunov conditions Behavior after large disturbances
Does convergence hold globally? Invariant domain and global argument with quantified initial conditions Whether real actors implement that process
Does software compute an equilibrium? Verified feasibility, optimality, residuals and complementarity at stated tolerances Historical price adjustment and economic identification
Does an institution attain a clearing allocation? A defined rule and appropriate observed order, trade, price and timing data Distributional valuation and external validity

The original analysis in this monograph consists of placing the same economic object under each of these questions, constructing explicit elementary comparisons, and following their implications for scientific practice. In particular, the fully specified two-household example establishes an equilibrium, verifies it algebraically, proves convergence for one continuous rule, and shows that a change in discrete implementation can reverse stability while preserving the solution. The local two-dimensional fields establish that an invertible Jacobian and a unique stationary point are compatible with attraction, divergence, and exact cycles. The moving-equilibrium bound establishes that nonzero tracking error can coexist with attractive frozen-fundamentals dynamics. Together these calculations clarify why apparent disagreement about equilibrium may reflect a change of analytical question rather than contradictory evidence.

The result is a methodological synthesis, not a proposed new equilibrium theory. All of the mathematical ingredients have established antecedents. Their derivations and application to the interpretation of economic research are supplied here as self-contained analytic illustrations. No priority claim is made for the examples, no new theorem of competitive realizability is asserted, and no laboratory study is re-estimated.

Several scientific problems remain particularly important. First, institutionally plausible adjustment rules should be compared at equal information and incentive costs rather than solely by sufficient convergence conditions. Second, identification of aggregate excess-demand derivatives should incorporate shifts of wealth and the endogeneity of order submission. Third, global stability analysis should address the interaction of admissible boundaries with actual trading and settlement constraints. Fourth, computational work should report distinct solutions and their conditioning when counterfactual conclusions depend on selection. Fifth, experimental designs should distinguish the role of market protocols from changes in the economic environment and track the distribution of trading gains.

These problems are tractable only when the target of each inquiry is specified. The existence of a solution is the foundation for equilibrium analysis. Determinacy, adjustment, computation, observation, and welfare are layers of inquiry that give this foundation empirical and institutional significance.

Conclusion

Competitive equilibrium establishes the joint compatibility of individually optimal choices and social feasibility at a relative price system. Modern existence theorems show that such compatibility can be guaranteed under precise conditions. The analysis of regular economies explains how locally isolated equilibrium branches can occur even in the presence of global multiplicity. The aggregate excess-demand literature establishes the freedom that survives individual optimization, and instability examples demonstrate the importance of explicit conditions on price motion.

The core findings are that stability belongs to a process specified together with an economic environment; that local regularity does not determine the signs of the dynamic eigenvalues; that the global behavior of a system depends on the domain, boundaries, and possible multiplicity of equilibria; and that equilibrium computation supplies a consistent numerical solution rather than independent evidence of historical attainment. A complete scientific evaluation therefore relates the formal equilibrium conditions to the institution that processes information and transactions.

Analytical demonstrations in the monograph establish these points transparently. The two-household exchange economy gives an exact stable continuous process and a range of discrete adjustment outcomes. The constructed linear fields explain multiple possible stability classifications at isolated regular solutions. The tracking calculation distinguishes adjustment failure from persistent changes in the equilibrium target. Experimental literature supplies qualitative evidence that market dynamics and selection can be studied directly through controlled trading institutions, while organized electricity markets illustrate why a clearing calculation and real-time balancing have distinct interpretations.

The original contribution is the integrated interpretation of the different evidentiary requirements and the self-contained comparison of mathematical, computational, and institutional questions. The limitations are finite-dimensional stylized models, selective primary and abstract-level verification, and the absence of original data or cross-country estimation. The study establishes a rigorous basis for further research on the existence, stability, and empirical adjustment of economic equilibria.

References

  • Arrow, K. J., & Debreu, G. (1954). Existence of an equilibrium for a competitive economy. Econometrica, 22(3), 265–290. DOI: 10.2307/1907353. Source
  • Arrow, K. J., Block, H. D., & Hurwicz, L. (1959). On the stability of the competitive equilibrium, II. Econometrica, 27(1), 82–109. DOI: 10.2307/1907779. Source
  • Scarf, H. E. (1960). Some examples of global instability of the competitive equilibrium. International Economic Review, 1(3), 157–172. Source
  • Debreu, G. (1970). Economies with a finite set of equilibria. Econometrica, 38(3), 387–392. DOI: 10.2307/1909545. Source
  • Sonnenschein, H. (1973). Do Walras' identity and continuity characterize the class of community excess demand functions? Journal of Economic Theory, 6(4), 345–354. DOI: 10.1016/0022-0531(73)90066-5. Source
  • Mantel, R. R. (1974). On the characterization of aggregate excess demand. Journal of Economic Theory, 7(3), 348–353. DOI: 10.1016/0022-0531(74)90100-8. Source
  • Debreu, G. (1974). Excess demand functions. Journal of Mathematical Economics, 1(1), 15–21. DOI: 10.1016/0304-4068(74)90032-9. Source
  • McFadden, D., Mas-Colell, A., Mantel, R., & Richter, M. K. (1974). A characterization of community excess demand functions. Journal of Economic Theory, 9(4), 361–374. Source
  • Scarf, H. E., with Hansen, T. (1973). The computation of economic equilibria. Cowles Foundation Monograph 24. Yale University Press. Source
  • Anderson, C. M., Plott, C. R., Shimomura, K.-I., & Granat, S. (2004). Global instability in experimental general equilibrium: The Scarf example. Journal of Economic Theory, 115(2), 209–249. Source
  • Crockett, S., Oprea, R., & Plott, C. R. (2011). Extreme Walrasian dynamics: The Gale example in the lab. American Economic Review, 101(7), 3196–3220. DOI: 10.1257/aer.101.7.3196. Source
  • Goeree, J. K., & Lindsay, L. (2016). Market design and the stability of general equilibrium. Journal of Economic Theory, 165, 37–68. DOI: 10.1016/j.jet.2016.04.004. Source
  • Gillen, B. J., Hirota, M., Hsu, M., Plott, C. R., & Rogers, B. W. (2021). Divergence and convergence in Scarf cycle environments: Experiments and predictability in the dynamics of general equilibrium systems. Economic Theory, 71(3), 1033–1084. DOI: 10.1007/s00199-020-01265-1. Source
  • Herings, P. J. J. (2024). Globally and universally convergent price adjustment processes. Journal of Mathematical Economics, 113, 103007. DOI: 10.1016/j.jmateco.2024.103007. Source
  • Federal Energy Regulatory Commission. (n.d.). Energy markets. U.S. Federal Energy Regulatory Commission. Accessed November 17, 2026. Source