When cooperation is beneficial to all agents

Abstract: Within a general semimartingale framework, we study the relationship between collective market efficiency and individual rationality. We derive a necessary and sufficient condition for the existence of (possibly zero-sum) exchanges among agents that strictly increase their indirect utilities and characterize this condition in terms of the compatibility between agents' preferences and collective pricing measures. The framework applies to both continuous- and discrete-time models and clarifies when cooperation leads to a strict improvement in each participating agent's indirect utility.

• The paper introduces a novel framework establishing when risk-sharing exchanges strictly improve agents' indirect utilities.
• It employs concepts of collective arbitrage and dual pricing measures to characterize beneficial cooperation in heterogeneous segmented markets.
• The analysis informs practical designs of risk-sharing contracts and regulatory policies by contrasting single-agent and collective market pricing.

Conditions for Universal Agent-Benefiting Cooperation

Introduction

This work rigorously analyzes the interplay between collective efficiency and individual rationality in segmented financial markets with multiple agents, each operating under potentially distinct market access, information sets, subjective probability measures, and utility functions. The focus is on specifying when risk exchanges (potentially zero-sum) not representable in any single agent's market lead to strictly higher indirect utility for all participants, thereby rendering cooperation universally individually rational.

The framework departs from and generalizes classical no-arbitrage and asset pricing results. The authors introduce the notions of Collective Arbitrage (CA) and No Collective Arbitrage (NCA)/No Collective Free Lunch (NCFL), establishing conditions under which coordinated action (risk-sharing by admissible exchanges $Y$ with $\sum_i Y^i = 0$ a.s.) creates Pareto improvements not accessible to solitary agents.

Formalization of the Problem

Segmented Market Model

Agents $i = 1, ..., N$ access individual asset sets with $\mathbb{R}^{d_i}$-valued processes $S^i$, operate under possibly heterogeneous filtrations $\mathbb{F}^i$, probability measures $P^i$, and utility functions $u^i$. Each agent faces their own terminal self-financing wealth cone $\mathcal{K}_i$. The set of exchanges $Y$ is any convex cone satisfying $\sum_i Y^i = 0$0 and typically consists of zero-sum reallocations, possibly constrained.

Collective Arbitrage Concepts

• Collective Arbitrage (CA): Existence of exchanges and admissible payoffs $\sum_i Y^i = 0$1 such that terminal payoffs are a.s. nonnegative for all, with strict positiveness on a set of positive measure for at least one agent.
• No Collective Arbitrage (NCA): Absence of CA; formalized via the aggregate value condition under agent-specific pricing measures.
• No Collective Free Lunch (NCFL): The analogous closure-based property in continuous time, ruling out weak-$\sum_i Y^i = 0$2 limits of free lunches.

Indirect Utility and Beneficial Exchanges

For random endowment $\sum_i Y^i = 0$3, the indirect utility for agent $\sum_i Y^i = 0$4 is

$\sum_i Y^i = 0$5

An exchange $\sum_i Y^i = 0$6 is beneficial if $\sum_i Y^i = 0$7 for all $\sum_i Y^i = 0$8, and strict for at least one $\sum_i Y^i = 0$9; strictly beneficial if strict for all.

Main Theorem and Characterization

The core technical result (Theorem 1) establishes a necessary and sufficient condition for the existence of (strictly) beneficial exchanges in terms of the relationship between two central objects:

• The set of collective separating measures $i = 1, ..., N$0 (vectors $i = 1, ..., N$1 giving at most zero aggregate value to any exchange in $i = 1, ..., N$2).
• The minimax vector $i = 1, ..., N$3, where each $i = 1, ..., N$4 minimizes the convex dual formulation of $i = 1, ..., N$5.

Equivalence:

Beneficial exchanges in $i = 1, ..., N$6 exist if and only if $i = 1, ..., N$7, i.e., there exists $i = 1, ..., N$8 with $i = 1, ..., N$9.

Interpretation

Thus, the existence of universally beneficial cooperation is precisely governed by a compatibility condition:

• If agents’ pricing functionals (as determined by their risk preferences and endowment) are not collectively consistent with the market’s collective pricing cone, strict improvement for all via cooperation is enabled.
• If the vector of individual preference-induced pricing measures is a collective separating measure, no strictly beneficial $\mathbb{R}^{d_i}$0 exists.

Implications and Theoretical Insights

Relation to Arbitrage and Market Completeness

• The presence of Collective Arbitrage (in either discrete or continuous time) guarantees the existence of strictly beneficial exchanges: arbitrageability implies the possibility of universally individually rational risk sharing.
• However, the converse does not hold: even under NCA/NCFL (i.e., in collectively arbitrage-free markets), beneficial $\mathbb{R}^{d_i}$1 often exist unless preferences are "aligned" with market structure.

Market completeness at the individual level eliminates all possible beneficial exchanges under NCA: if each agent faces a complete arbitrage-free market, there is no scope for universally beneficial cooperation.

Connection to Pareto Optimality

Beneficial exchanges correspond to Pareto improvements (especially when $\mathbb{R}^{d_i}$2 is the zero-sum cone). The authors give a variational characterization and show that, under appropriate compactness/continuity, optimizers over positive linear combinations of indirect utilities are Pareto optimal.

Distinction from Representative-Agent View

The social planner/representative-agent arbitrage condition does not coincide with NCA in general unless there is informational and asset-alignment across agents. Multi-agent structure generates risk-sharing possibilities that are not reducible to aggregate market feasibility.

Explicit Examples

• The paper provides discrete-time and continuous-time examples, including complete algebraic constructions, verifying the theorem.
• In the two-agent case, if subjective probabilities differ on some event $\mathbb{R}^{d_i}$3, the zero-sum exchange $\mathbb{R}^{d_i}$4 leads to a strict aggregate gain in expected utility provided the agents' minimax measures disagree on $\mathbb{R}^{d_i}$5.

Practical and Theoretical Ramifications

This theoretical structure is directly relevant to markets featuring frictions, information asymmetry, segmentation, and heterogeneity in risk preferences and beliefs.

Practical implications include:

• Design of risk-sharing institutions and contracts that extract mutual gains unavailable to isolated trading.
• Interpretation of financial regulation: preventing certain classes of collective arbitrage does not eliminate all beneficial risk-sharing activity.
• Model-based testing for robust risk-sharing opportunities given observed market incompleteness and agent heterogeneity.

Theoretical consequences include:

• Precise characterization of when coordination is both rational and strictly advantageous for all participants.
• The necessity to go beyond single-agent pricing theorems in settings with segmentation and randomness in admissible trades.

Future research may generalize to dynamic mechanisms, stochastic coalition formation, principal-agent relationships, or more intricate frictions.

Conclusion

The paper provides a mathematically complete and conceptually clear solution to the characterization of universally beneficial cooperation among heterogeneous agents in segmented markets. The central result, hinging on the duality between agents' preference-induced pricing and the market's collective separating pricing, tightly demarcates when cooperation yields strict improvements for all agents. This generalizes classic no-arbitrage and welfare theorems in a direction suitable for modern multi-agent and segmented-market applications, yielding insights relevant to financial economics, mechanism design, and multi-agent system theory.