Transferable-Utility Games

• Transferable-Utility Games are formal cooperative models where any coalition can freely redistribute its joint value, defined by a characteristic function.
• They use solution concepts such as the core, Shapley value, and Gamma value to ensure fair and stable allocation in diverse market and network settings.
• Applications include matching markets, cost-sharing, and energy systems, with research extending the theory to dynamic, uncertain, and network-constrained environments.

A transferable-utility (TU) game is a formal model in cooperative and matching markets in which a group of agents (players) can freely redistribute (“transfer”) any jointly generated value among themselves. In such games, the worth of each coalition is described by a characteristic function, and redistribution schemes—ranging from classical allocations like the Shapley value and the core to more recent proposals tailored to specific agent networks or market structures—are central topics. Transferability of utility underpins a highly tractable theory for stable payoff division, leads to powerful duality results (e.g., in assignment games), and supports a diversity of solution concepts crucial to economic design and algorithmic game theory.

1. Formal Models and Structures

At the core of a TU game is a finite player set $N = \{1, \dots, n\}$ and a characteristic function $v : 2^N \to \mathbb{R}$ with $v(\emptyset) = 0$. The function $v(S)$ represents the total utility that coalition $S$ can ensure for itself, assuming it secedes from the grand coalition. The essential property of transferability is that, for any victorious coalition $S$, the sum $v(S)$ can be divided among its members in any fashion: if $x(S) = v(S)$, the allocation $x \in \mathbb{R}^N$ is feasible, provided $x_i = 0$ for $v : 2^N \to \mathbb{R}$0.

The class of TU games subsumes cost-sharing games (where the focus is on distributing costs), standard surplus-sharing games, assignment markets, and profit-sharing scenarios with or without additional structure (e.g., a priori unions, network constraints, uncertainty, or dynamic formation processes) (Besson-Niebles et al., 1 Dec 2025, Babaioff et al., 2019, Choudhury et al., 2022, Alonso-Meijide et al., 2024).

Variants of these basic models include:

• Matching games with transferability, e.g., the generalization to linearly-transferable-utility (LTU) where the rate of transfer within pairs may itself be type-dependent (Galichon et al., 2024).
• Partition-function games, which assign surplus not only to coalitions but to coalition structures/partitions, requiring lattice-theoretic techniques (Rossi, 2018).
• Robust or dynamic games, where coalition values are drawn from (possibly time-varying) uncertainty sets and bargaining is distributed over time-varying networks (Raja et al., 2020, Bauso et al., 2011).

2. Solution Concepts and Stability

The main criterion for stability of an allocation $v : 2^N \to \mathbb{R}$1 in TU games is the core. The core is defined as the set of payoff vectors that are both efficient (budget-balanced) and prevent any coalition from benefiting by leaving the grand coalition: $v : 2^N \to \mathbb{R}$2 where $v : 2^N \to \mathbb{R}$3. The core may be empty (notably in subadditive games such as large traveling salesman games) and its feasibility is characterized by the balancedness of the game via the Bondareva–Shapley theorem (Besson-Niebles et al., 1 Dec 2025, Kroupa et al., 2018).

When the core is empty or too small, relaxation concepts arise:

• The semicore, imposing only constraints on singleton and “marginal” coalitions (Besson-Niebles et al., 1 Dec 2025).
• The $v : 2^N \to \mathbb{R}$4-core, allowing excess up to $v : 2^N \to \mathbb{R}$5 on coalitional constraints (Besson-Niebles et al., 1 Dec 2025).
• The cost of stability, measuring the minimal subsidy for which the core becomes nonempty (Besson-Niebles et al., 1 Dec 2025).
• Robust cores for families of games under value uncertainty (Raja et al., 2020).

Special classes, such as totally balanced games (where every subgame has a nonempty core), have rich polyhedral descriptions essential for understanding core-structure in markets (Kroupa et al., 2018).

The following table summarizes core-related notions:

Solution Concept	Efficiency?	Full Stability?	Existence Guarantee
Core	Yes	All coalitions	Balancedness
Semicore	Yes	1 and n-1 coalitions	Often even when core empty
$v : 2^N \to \mathbb{R}$6-core	Yes	All, approx.	For all $v : 2^N \to \mathbb{R}$7
Cost of stability	Subsidized	All coalitions	Always, with enough subsidy
Robust core	Yes	All models in family	Intersected core nonempty

3. Classic and Modern Value Allocations

Several value solutions prescribe explicit payoff allocations for TU games:

• Shapley value: The archetype for marginalist payoff sharing. Each player receives their average marginal contribution across all coalition orders:

$v : 2^N \to \mathbb{R}$8

Satisfies efficiency, symmetry, linearity, and the null-player axiom. Generalizations include extensions to digraph games, k-Solidarity-Egalitarian values (k-SED), and the Gately and Γ (“Gamma”) values, each modifying marginalism with solidarity or necessary-player axioms (Khatri, 2017, Choudhury et al., 2022, Gilles et al., 2022, Gonçalves-Dosantos et al., 2024).

• Equal Division and Egalitarian Surplus Division: These split total surplus or residual surplus after individual values equally, and can be axiomatized for games with a priori unions (Alonso-Meijide et al., 2024).
• Gately value: Allocates the surplus so as to minimize the maximum “propensity to disrupt” (threat of defection), providing a unique balancing allocation under regularity conditions (Gilles et al., 2022).
• Γ-value (Gamma value): Emphasizes necessary players and per-capita coalition averages, filling a role intermediate between Shapley and equal division, with further generalization to coalition structures (Gonçalves-Dosantos et al., 2024).
• Partition Lattice Values: In games on partitions, “atom-based” Möbius inversions deliver chain-uniform and size-uniform values for partition-function-based TU games, generalizing Shapley-style reasoning to coalition-structure settings (Rossi, 2018).

These solutions are motivated by a variety of fairness, efficiency, and strategic axioms and often coincide or differ systematically depending on particular structural features of the game (e.g., presence/absence of necessary players or coalition symmetries).

4. Algorithmic and Computational Aspects

Computational analysis of TU games spans core membership, value calculation, and stable matching computation in both static and dynamic models:

• Core membership and emptiness: In subadditive combinatorial games (such as TSP-based games), core membership is coNP-hard, and even existence must often be checked using the ellipsoid method with separation oracles (Besson-Niebles et al., 1 Dec 2025).
• Assignment and matching games: In classical matching with (fully) transferable utility, stable outcomes coincide with integer LP solutions or dual prices. The addition of middlemen maintains total balancedness and the assignment game equivalence, but no single “middleman-optimal” allocation generally exists (Atay et al., 2021).
• Linearly transferable utility (LTU): The LP structure collapses in LTU games; stable outcomes correspond to Nash equilibria of bimatrix “hide-and-seek” games. Computing such equilibria is PPAD-complete in general (Galichon et al., 2024).
• Distributed computation and bargaining: In robust or dynamic coalitional games, consensus-based protocols allow distributed agents to iteratively negotiate towards the (robust) core using only local communication, with convergence under mild connectivity and operator-theoretic conditions (Raja et al., 2020, Bauso et al., 2011).
• Exact computation of allocation rules: For classic solution concepts—Shapley, k-SED, Gately, Gamma values—closed-form or tractable algorithms exist, with increased complexity as constraints, network/partition structure, or information sets become richer (Choudhury et al., 2022, Gilles et al., 2022, Alonso-Meijide et al., 2024).

5. Generalizations and Extensions

Research continues to extend the paradigm of TU games on several axes:

• Games under uncertainty: Robust and dynamic TU games address coalitional values subject to bounded or stochastic fluctuations. Robust cores and average-game protocols ensure stability under value ambiguity (Raja et al., 2020, Bauso et al., 2011).
• Network and partition-function games: Partition-function-form (PFF) games and global TU games on lattices of partitions capture coalition-structure-dependent surpluses. Atom-based values and new core criteria substitute for traditional core logic, and supermodularity is not sufficient for core nonemptiness in the partition setting (Rossi, 2018).
• Approximate stability: Infeasibility of full core stability in large or subadditive games (e.g., TSGs) motivates a range of approximate schemes: semicore, cost of stability, $v : 2^N \to \mathbb{R}$9-core, and relaxations parameterized by coalition size or marginal constraints. These are especially critical in practical markets such as collaborative transportation or assignment systems (Besson-Niebles et al., 1 Dec 2025).
• A priori unions and coalition structures: Values and fairness rules are extended to situations with fixed grouping structures, such as departments, floors, or division in organizations. Multi-level allocation mechanisms preserve fairness both within and across these unions (Alonso-Meijide et al., 2024, Gonçalves-Dosantos et al., 2024).

6. Applications and Impact

Transferable-utility game models are standard in matching markets (marriage, labor, or school assignment), cost-sharing (e.g., transportation or facility location), profit-sharing, communication networks, and distributed resource allocation:

• Matching and assignment markets: TU structure underlies competitive equilibrium in assignment markets and the classical realization that market clearing prices correspond to dual variables in optimum assignment LPs. Middlemen introduce rich core structures without loss of balancedness (Atay et al., 2021).
• Collaborative transportation and facility allocation: Stability of coalition formation (e.g., among producers in short food supply chains, or building owners dividing elevator costs) critically depends on precise allocation rules and their computational practicality; approximate stability and robust solutions mitigate difficulties arising from empty cores (Besson-Niebles et al., 1 Dec 2025, Gonçalves-Dosantos et al., 2024).
• Energy storage and smart power grids: Distributed, robust, and time-varying TU games provide a viable model for prosumer-based cooperation in energy systems, with consensus protocols enabling stable sharing of cooperative benefits under uncertainty (Raja et al., 2020).
• Fair division and rental-sharing: Recent algorithms for profit-sharing and allocation combining efficiency, fairness, and decomposability fill gaps left by standard concepts (Shapley, Kalai–Smorodinsky, Envy-Freeness) and admit strongly polynomial algorithms when submodular (Babaioff et al., 2019).

The analytic machinery of TU games remains foundational for both cooperative game theory and its computational-economic applications, particularly as new domains motivate increasingly nuanced generalizations and practical considerations.