📈 economics

The Core in a Distributional Economy

This paper refutes the notion that distributional descriptions of economies are insufficient for core analysis by demonstrating that blocking coalitions can be identified solely through distributions and providing a purely distributional proof of the classical core-equivalence theorem.

Original authors: Michael Greinecker, Konrad Podczeck

Published 2026-04-29

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Michael Greinecker, Konrad Podczeck

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine an economy not as a room full of specific people with names, faces, and unique histories, but as a giant, swirling cloud of characteristics. In this cloud, you don't see "Alice the baker" or "Bob the teacher." Instead, you see a statistical map: "30% of the cloud likes bread, 20% likes shoes, and everyone has a certain amount of flour or leather to start with."

For decades, economists believed that to understand how this cloud behaves—specifically, whether a group of people could gang up to make themselves better off (a concept called the Core)—you had to know who the specific individuals were. They thought the "cloud" view was too blurry to see the details of who was blocking whom.

The Big Idea: The Cloud is Enough
Michael Greinecker and Konrad Podczeck's paper argues that this is wrong. You don't need to know the names of the people in the cloud. You can solve the puzzle using only the statistics of the cloud.

Here is how they do it, using some creative analogies:

1. The "Ghost" Problem vs. The "Super-Cloud"

Usually, if you look at a distribution (a cloud), you lose track of individual identities. It's like looking at a bag of mixed jellybeans and knowing there are 50 red ones and 50 blue ones, but not knowing which specific bean is which.

The authors introduce a mathematical concept they call a "Super-Cloud" (or a superatomless space). Think of this not as a bag of distinct beans, but as a perfectly smooth, infinitely divisible liquid. In this liquid, you can scoop out any fraction of the liquid you want, no matter how small or how specific the mixture, and it will always behave exactly like a real group of people.

Because this "Super-Cloud" is so mathematically rich, any "group" you can imagine forming in the statistics (the distribution) actually corresponds to a real, physical group of agents in the background. This solves the problem of "who is in the coalition?" without needing to name them.

2. The "Gang Up" (Blocking Coalitions)

In economics, a "blocking coalition" is a group of people who say, "We don't like the current deal. If we just trade amongst ourselves, we can all be better off."

The Old Way: You had to find a specific list of names (Alice, Bob, and Charlie) to prove they could block the deal.
The New Way: You just look at the distribution. You ask: "Is there a shape of a group in our statistical cloud that, if they traded amongst themselves, would improve their lot?"

The paper proves that if such a "shape" exists in the statistics, a real group must exist to do the blocking. You don't need to find the names; the statistics guarantee the group's existence.

3. The "Perfect Shuffle" (Pareto Efficiency)

To check if an economy is efficient (meaning no one can be made better off without hurting someone else), you usually have to compare Person A's new life to Person A's old life.

The authors show that in their "Super-Cloud," you can use a concept called Stochastic Dominance. Imagine you have two different ways to distribute the jellybeans. Instead of tracking every single bean, you just check: "In the new distribution, does the 'top half' of the crowd get better beans than the 'top half' in the old distribution?"

They prove that if the statistics show the new distribution is "better" in this statistical sense, then there is a way to shuffle the actual people so that everyone is indeed better off. It's like proving a deck of cards is better shuffled just by looking at the pattern of suits, without needing to track the Ace of Spades specifically.

4. The Grand Conclusion: The Core Equivalence

The most famous result in this field is the Core-Equivalence Theorem. It basically says: "In a huge economy, the only deals that survive (the Core) are the ones that look like a perfect market price system (Walrasian Equilibrium)."

Historically, proving this required a complex, two-step process:

Define the economy with specific people.
Prove the theorem.
Then try to translate it back to a distribution.

The authors do the reverse. They prove the theorem directly using the distribution. They show that the "Core" and the "Market Price" are the same thing even if you never define a single individual.

The Takeaway
Think of the economy as a massive, complex weather system. Previously, economists thought you couldn't predict a storm (the Core) without tracking every single water molecule (the individual agents).

Greinecker and Podczeck show that you don't need to track the molecules. If you understand the pressure, temperature, and humidity patterns (the distribution) correctly, you can predict the storm perfectly. The "individuals" are just the water molecules; the "economy" is the weather. You can study the weather without counting the drops.

What this means for the paper's scope:
The authors demonstrate that their mathematical tools work for:

General Equilibrium Theory: Proving market stability without individual names.
Large Matching Markets: Matching people (like doctors to hospitals) in huge systems using only statistics.
Shapley-value analogs: Calculating "fair shares" in atomless (infinitely divisible) economies.

They do not claim these tools apply to clinical settings or specific future technologies; their contribution is strictly a foundational mathematical proof that the "distribution-only" view is just as powerful as the "individual" view for understanding how large economies function.

1. Problem Statement

Traditional general equilibrium theory defines economies by assigning specific characteristics (preferences, endowments) to a continuum of individual agents. While competitive equilibria can be defined purely in terms of the distribution of these characteristics (as shown by Hart, Hildenbrand, and Kohlberg), the concept of the core has resisted such a formulation.

The core relies on the notion of blocking coalitions, which traditionally requires an individualistic description of agents to determine:

Identity: Which specific agents are in a coalition.
Count: How many agents of a specific type exist to form a blocking group.
Feasibility: Whether the coalition can redistribute its aggregate endowment to make members better off.

Critics (e.g., Hart, Hildenbrand, Kohlberg; Mas-Colell) have argued that a distributional description is too "sparse" for core analysis because it loses information about agent identity and the specific number of agents per type. This paper challenges that assertion, asking: Can the core and blocking coalitions be defined and analyzed using only the distribution of characteristics, without reference to individual agents?

2. Methodology

The authors employ advanced measure-theoretic tools to bridge the gap between individualistic and distributional economies.

Distributional Allocations: They define an economy not as a map from agents to characteristics, but as a probability measure $\mu$ on the space of characteristics (preferences $\times$ endowments). An allocation is a joint distribution of characteristics and commodity bundles.
Generalized Coalitions: They interpret coalitions not as sets of agents, but as subpopulations (measures $\nu$ such that $\nu \leq \mu$ ). This aligns with concepts like "ideal sets" (Aumann & Shapley), "fuzzy coalitions" (Aubin), and "convex sets of agents."
Superatomless Spaces: A crucial mathematical innovation is the reliance on superatomless probability spaces.
- A space is atomless if it has no indivisible atoms.
- A space is superatomless (or saturated) if it is "rich" enough to represent any conceivable joint distribution of characteristics and allocations. Formally, a space $(T, \Sigma, \nu)$ is superatomless if $L^1(A, \nu)$ is non-separable for all $A$ with $\nu(A) > 0$ .
- This property ensures that any distributional coalition (subpopulation) can be exactly represented by a measurable set of agents in the underlying space.
Stochastic Dominance & Strassen's Theorem: To define Pareto efficiency and blocking without joint distributions, they utilize first-order stochastic dominance (Reif & Wiesmeth) and Strassen's Theorem (and its extension by Lindvall) on stochastic pre-orderings. This allows them to construct joint distributions from marginal distributions that satisfy specific ordering conditions.
Lyapunov's Theorem for Young Measures: They use Balder's (2000) extension of Lyapunov's theorem to handle the convexification of sets of allocations, proving the existence of blocking allocations even when the underlying space is merely atomless (not necessarily superatomless) for the existence of some blocking coalition.

3. Key Contributions

A. Characterization of Pareto Efficiency

The paper demonstrates that Pareto efficiency can be fully characterized in distributional terms.

Result: An allocation is Pareto efficient if and only if there is no other distributional allocation that stochastically dominates it in the sense of Reif and Wiesmeth.
Significance: This resolves the issue of "lost identity." Even without knowing which agent is better off, the joint distribution of characteristics and bundles allows for a complete ranking of allocations.

B. Definition of the Distributional Core

The authors define a distributional core where an allocation is blocked if there exists a subpopulation (a measure $\mu'$ dominated by the economy's distribution) that can redistribute its endowment to make its members strictly better off.

Theorem 4 (Superatomless Case): If the underlying agent space is superatomless, every distributional blocking coalition corresponds exactly to a classical individualistic coalition.
Theorem 5 (Atomless Case): Even if the space is only atomless (not superatomless), the existence of a distributional blocking coalition implies the existence of a classical blocking coalition. This is a critical result: one does not need the "rich" space to prove that a distributional economy has a non-empty core; the distributional condition is sufficient to guarantee the existence of a real-world blocking coalition.

C. Core-Equivalence Theorem

The paper provides a purely distributional proof of the Core-Equivalence Theorem (Theorem 7).

Statement: In a distributional economy with continuous, irreflexive, strictly monotone preferences and a strictly positive aggregate endowment, an allocation is in the core if and only if it is a Walrasian (competitive) equilibrium.
Novelty: Unlike previous proofs (e.g., Hildenbrand, 1982), this proof does not rely on an ancillary individualistic representation. It operates entirely within the distributional framework.
Implication: The classical core-equivalence theorem for individualistic economies with atomless spaces is derived as a direct corollary.

4. Key Results Summary

Concept	Traditional View	Distributional View (This Paper)
Economy	Map $E: T \to P \times E$	Measure $\mu$ on $P \times E$
Coalition	Measurable set $C \subset T$	Sub-measure $\nu \leq \mu$
Blocking	$f'(t) \succ_t f(t)$ for $t \in C$	Stochastic dominance of sub-population
Pareto Efficiency	No $f'$ dominates $f$	No $\alpha'$ dominates $\alpha$ (via Strassen)
Core Equivalence	Requires individualistic proof	Proven purely distributionally; implies individualistic case
Mathematical Tool	Lyapunov's Theorem	Lyapunov for Young Measures + Superatomlessness

5. Significance and Applications

Theoretical Unification: The paper unifies various "generalized coalition" concepts (ideal sets, fuzzy coalitions, subpopulations) under a single distributional framework. It shows these are merely different mathematical representations of the same underlying economic reality: a redistribution of a subpopulation of types.
Parsimony: It proves that the "individualistic" description is not necessary for core analysis. The distribution of characteristics alone contains all economically relevant information for determining stability and equilibrium.
Robustness: By showing that distributional blocking implies individualistic blocking even in merely atomless spaces, the authors demonstrate that the distributional approach is robust and not an artifact of overly strong mathematical assumptions (like superatomlessness).
Broader Applications: The authors note that these methods extend beyond exchange economies to:
- Large matching markets.
- Shapley-value analogs for atomless economies.
- Continuum games.

Conclusion

Greinecker and Podczeck successfully dismantle the long-held belief that the core requires an individualistic description of agents. By leveraging the properties of superatomless spaces and stochastic dominance, they construct a rigorous, purely distributional theory of the core. This not only simplifies the theoretical framework of large economies but also provides a more powerful tool for analyzing economies where individual identities are irrelevant or unobservable, confirming that the distribution of types is sufficient to determine market stability.