Is the "Yard-Sale" wealth condensation theorem an economic insight or a mathematical artifact of its transaction rule?

Quoted from Forex.com.bd-Editorial External question — Economics Stack Exchange Author: Advaita Source score (net votes, not local likes): 0 Original post: https://economics.stackexchange.com/questions/61115 License: CC BY-SA 4.0 — https://creativecommons.org/licenses/by-sa/4.0/ Adaptation: HTML converted to plain text; contact email addresses removed. ( NOTE:I USED AI TO REFINE THE SEMANTICS OF MY QUESTION https://math.stackexchange.com/questions/5141882/is-the-yard-sale-wealth-condensation-theorem-an-economic-insight-or-a-mathemat In econophysicist literature— particularly Börgers and Greengard (2024 and 2026) (please take a look at Börgers, C., & Greengard, C. (2026). Extreme wealth inequality from randomness. The Mathematical Gazette, 1–12. https://doi.org/10.1080/00255572.2025.2544451 )—the "Yard-Sale Model" is used to argue that unbiased random markets inherently cause absolute wealth concentration where $\lim_{k \to \infty} R_k = 1$ almost surely. However, the model enforces an exogenous transaction rule where the traded wealth amount is always a fixed fraction $b$ of the poorer agent's asset pool. Because this forces the poorer agent into a multiplicative random walk of $1 \pm b$ , their wealth is mathematically dragged toward zero over time. From an economic perspective, is the Yard-Sale Convergence Theorem providing a meaningful insight into market dynamics, or is it merely a mathematical artifact of forcing lower-wealth agents to repeatedly wager an invariant percentage of their entire net worth? The model outlines a closed economy of $n \ge 2$ agents whose wealth vectors exist on an open standard $(n-1)$ -simplex, meaning total wealth is conserved. The simplex is defined as: $$\Delta^{n-1} = \left\{ (x_1, \dots, x_n) \in \mathbb{R}^n : 0 < x_i < 1 \text{ for all } i \text{ and } \sum_{i=1}^n x_i = 1 \right\}$$ In each discrete trade $k$ , a pair of agents $(I_k, J_k)$ is chosen uniformly at random, independent of previous choices. To track wealth-dependent dynamics, the model enforces a sorting mechanism: if the wealth of the first chosen agent is greater than the second, their indices are swapped. This guarantees that $I_k$ always denotes the poorer (or equal) trading partner before the trade occurs: $$X_{I_k}^{k-1} \le X_{J_k}^{k-1}$$ The amount of wealth exchanged is defined exogenously as $bX_{I_k}^{k-1}$ , where $b \in (0,1)$ is a fixed parameter. The direction of the transfer is decided by a fair coin flip $S_k \in \{-1,1\}$ with equal probability. The wealth update rules are given by: $$X_{I_k}^k = X_{I_k}^{k-1} - S_k b X_{I_k}^{k-1}$$ $$X_{J_k}^k = X_{J_k}^{k-1} + S_k b X_{I_k}^{k-1}$$ $$X_i^k = X_i^{k-1} \quad \text{for } i \notin \{I_k, J_k\}$$ As the authors note to build intuition, during these trades, the wealth of the less wealthy agent is multiplied by $1 \pm b$ . After $j$ trades, as long as the poorer agent remains the poorer, their wealth goes down by an approximate factor of $(1-b^2)^{j/2}$ . This geometric decay ensures that the minimum wealth $L_k \to 0$ almost surely, eventually collapsing the entire simplex into a single canonical basis vector where one agent holds all the wealth. While the paper elegantly proves this via martingale convergence and the Borel-Cantelli lemma, the underlying economic mechanism seems problematic. In microeconomic theory, exchange is driven by utility maximization, comparative advantage, or at least voluntary participation. In this model: No Micro-foundations: Agents do not optimize utility; they are exogenously forced to gamble. Asymmetric Risk Exposure: The richer agent wagers a negligible fraction of their wealth $\frac{bX_{I_k}^{k-1}}{X_{J_k}^{k-1}}$ , while the poorer agent is structurally forced to wager a massive, constant fraction $b$ of their entire net worth. Furthermore, the authors note that attempting to adjust this proof to handle a "pro-poor bias" (a wealth-acquired disadvantage) requires a highly restrictive condition implying that the bias must disappear in the limit. I have looked into how changing the transaction rule alters the martingale properties. If we modify the model so that the trade volume is an absolute constant amount $\epsilon$ (where $\epsilon < \min(X_i)$ ) rather than a percentage of the poorer agent's wealth $bX_{I_k}^{k-1}$ , the transaction engine shifts from a multiplicative random walk to an additive random walk on a simplex. While the discrete martingale property $E(X_i^k \mid X_i^{k-1}) = X_i^{k-1}$ still holds, the absolute wealth concentration theorem ( $\lim_{k \to \infty} R_k = 1$ ) breaks down because the variance no longer scales down symmetrically to crush the poorer agent's position. Furthermore, from a game-theoretic or behavioral standpoint, no rational agent approaching the lower boundary of wealth would voluntarily enter a $50/50$ coin-flip exchange where the stake is a fixed percentage of their entire livelihood. Are there any existing papers in economic literature that criticize this specific feature of kinetic exchange models? How do economists reconcile these "physical/statistical" wealth models with the reality that transaction sizes do not scale strictly to the total net worth of the poorest participant in a trade? --- Update: Handling the Boundary Condition ($X_i < \epsilon$) A sharp point was raised in the comments regarding how the modified additive model handles agents whose wealth drops below the constant transaction size $\epsilon$ . In a non-negative closed economy ( $X_i \ge 0$ ), shifting from a multiplicative walk to a flat additive walk requires defining a boundary condition at the lower limit. There are two standard stochastic approaches to this, and both highlight the structural fragility of the original Yard-Sale convergence: The Truncation / "No-Trade" Rule (Insulating Boundary): If we assume trades are aborted when the poorer agent cannot afford the flat "entry fee" of $\epsilon$ , the stake becomes: $$\Delta W = \epsilon \quad \text{if } X_{I_k}^{k-1} \ge \epsilon, \quad \text{otherwise } \Delta W = 0$$ This creates a hard mathematical safety net just above zero. Because an agent's wealth can never be dragged below $\epsilon$ , it becomes impossible for any agent's wealth to hit 0. Under this rule, the absolute wealth concentration theorem ( $\lim R_k = 1$ ) completely falls apart, and the system stabilizes into a fluctuating, non-monopolistic state. The Capped Stake / "All-In" Rule (Absorbing Boundary): If we instead cap the transaction size by whatever the poorer agent has left, the stake is defined dynamically as: $$\Delta W = \min(\epsilon, X_{I_k}^{k-1})$$ In the interior of the simplex, it behaves as an additive walk. However, if an agent's wealth drops below $\epsilon$ , they are forced into an "all-in" gamble. If they lose, their wealth hits exactly $0$ , making it an absorbing state. Under this rule, absolute wealth concentration still occurs because agents are systematically eliminated one by one at the boundary. The Core Critique: The value of this comparison is that it exposes why the Yard-Sale Model behaves the way it does. The original model's multiplicative rule ( $bX_{I_k}$ ) is applied universally across the entire interior of the simplex, creating a continuous geometric drag toward zero ( $1-b^2$ ) that operates on the poor everywhere , rather than a localized boundary risk.

Cancel quote