Balanced contributions, consistency, and value for games with externalities Journal: to arXiv

André Casajus, Yukihiko Funaki, Frank Huettner · 2025

Read Balanced contributions, consistency, and value for games with externalities Journal: to arXiv by André Casajus, Yukihiko Funaki, Frank Huettner (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2511-03145.txt.

Introduction

Andre Casajus, Yukihiko Funaki, and Frank Huettner extend the Shapley value, the classic fair-share rule of game theory, to games where outside parties affect what a group can earn. In such games with externalities, fairness and consistency need fresh definitions. Their 2025 paper builds on a restriction found in earlier work to define balanced contributions and two kinds of consistency for this harder setting.

They then show those properties characterize a generalized Shapley value proposed by Macho-Stadler and colleagues in 2007. The characterization mirrors the famous ones for the original Shapley value: adopt the listed fairness and consistency demands, and this value is the only rule left standing. Such results matter for dividing gains when coalitions cannot seal themselves off from rivals.

Catalog record

CREATORAndré Casajus, Yukihiko Funaki, Frank Huettner
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE47 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2511-03145

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Balanced contributions, consistency, and value for games with externalities Journal: to arXiv André Casajus, Yukihiko Funaki, Frank Huettner Abstract We consider fair and consistent extensions of the Shapley value for games with externalities. Based on the restriction identified by Casajus et al. (2024, Games Econ. Behavior 147, 88-146), we define balanced contributions, Sobolev’s consistency, and Hart and Mas-Colell’s consistency for games with externalities, and we show that these properties lead to characterizations of the generalization of the Shapley value introduced by Macho-Stadler et al. (2007, J. Econ. Theory 135, 339-356), that parallel important characterizations of the Shapley value. 2010 MSC 91A12, JEL: C71, D60 1 Introduction The question of how to divide jointly created value in cooperative games has been fundamentally influenced by the concept of the Shapley value. This solution concept was originally designed to capture a player’s value in a cooperative game with transferable utility (henceforth TU game), but its application nowadays transitions into seemingly unrelated domains such as statistics for identifying important variables ( 37 ; 59 ) or into machine learning for interpreting prediction models ( 40 ; 41 ) . The Shapley value applies under the presupposition that the worth of a coalition is independent of the coalition structure of the outside players. However, this assumption is challenged when external effects come into play, that is, when the actions of one coalition exert influence over another’s outcomes. To capture externalities, 66 introduced the framework of games with externalities (henceforth TUX games ). Crucially, in a TUX game, the worth of a coalition may depend on the coalitions formed by the outsiders, which makes it a generalization of TU games. TUX games are often used—but are not limited—to study coalition formation ( 30 ; 5 ) . Indeed, recently 57 capture covert networks with TUX games to evaluate agents’ importance in such networks, demonstrating that TUX games can lead to more powerful models and insights than TU games. Whereas the Shapley value probably is the most-prominent single-valued solution concept for TU games, there is a lack of clarity when it comes to solutions for TUX games. Indeed, many generalizations of the Shapley value to TUX games were proposed in the literature, usually motivated by characterizations in the spirit of Shapley’s original characterization based on additivity. 1 1 1 50 , 10 , 1 , 54 , 42 , 47 , 22 , 30 , and 60 introduce different (classes of) solutions. However, the Shapley value (and derivatives thereof) are often justified by axioms that involve varying player sets, such as the balanced contributions property ( 51 ) and consistency properties deriving from 33 and from 62 . Therefore, 22 and 12 argue that generalizations of the Shapley value should also inherit its properties that relate to varying player sets. After all, these properties not only lie at the heart of many fairness properties of the Shapley value, but also build the fundament for implementations via non-cooperative games (see, e.g., 31 , 42 , 45 , 11 ). 1.1 Novel axioms for games with externalities and characterizations of the MPW Solution In this paper, we define a notion of balanced contributions for games with externalities along the lines of 51 , which requires that the impact of removing player j on player i ’s payoff is the same as the impact of removing player i on player j ’s payoff. We demonstrate that this property together with efficiency is characteristic of the solution introduced by 25 and characterized by 42 , henceforth abbreviated as MPW solution , a generalization of the Shapley value (Theorem 4 ). Moreover, we investigate the consistency properties introduced by 33 and by 62 , which are well-known to be characteristic of the Shapley value together with efficiency and standardness for two-player games. We augment these consistency properties to TUX games . Our second main contribution are two novel characterizations of the MPW solution (Theorems 8 and 9 ). Generalizing 62 to games with externalities, we show that the generalized version of Sobolev’s consistency together with efficiency and standardness for two-player games is characteristic of the MPW solution. Generalizing 33 to games with externalities, we show that the generalized version of Hart and Mas-Colell’s consistency together with standardness for two-player games is characteristic of the MPW solution. In this sense, we contribute to the stream of literature that generalizes these consistency properties to broader frameworks with the intend to generalize the Shapley value (see 69 , 22 , or 70 for other such attempts). 1.2 Relation to the Literature Several motivations for studying balanced contributions and consistency arise from the literature. We will give a brief summary. 1.2.1 Balanced contributions property is a fairness property that is fruitful for derived axiomatizations Unlike additivity, the balanced contributions property is a fairness statement ( 49 ) . It establishes an equitable relationship between players’ contributions to each other. This property has proven instrumental in characterizing various extensions of the Shapley value (e.g., 39 , 27 , 35 , and 68 ). The balanced contributions property was also key for motivating specific applications of the Shapley value, including less traditional applications in machine learning ( 18 ) , as well as applications in more traditional realms. Recent examples involve network games ( 28 ) , revenue sharing ( 8 ) , and group decision models ( 48 ) . These applications demonstrate how balanced contributions can effectively embody core fairness principles when adapted to particular operational contexts. It further suggests that extending this axiom to games with externalities can lead to a plethora of derivative insights. 1.2.2 Merits of consistency properties Exploring the consistency properties within cooperative game theory serves a tripartite purpose ( 21 ) . First, it enables a clear differentiation between various solution concepts, offering a nuanced perspective on their unique properties and conditions of application. For comparison, 3 provide a characterization of the prenucleolus for TUX games based on a consistency property generalizing the notion of the reduced game introduced by 19 . Second, exploring consistency properties fosters the theoretical evolution of these solution concepts, deepening our understanding of their underpinnings and potential for refinement. This is, for example, important for the development of mechanisms that implement a solution concept by a non-cooperative game. To this end, it is particularly useful to know about ways to pay out players and reduce the size or complexity of a game without changing the payoffs of the remaining players. Prominent examples for mechanisms that implement the Shapley value and that stepwise reduce the number of players are given by 42 , 45 , and 46 ; implementations of derivatives of the Shapley value along such lines are given by, for example, 9 , 34 , and 7 . Lastly, the consistency property may be useful in order to determine coherent solutions for realistic problems, that is, when the game derives from specific applications. Important examples for which consistency properties of TU games boil down to meaningful properties for the specific application include airport problems ( 38 ; 56 ) ; bankruptcy problems ( 52 ; 4 ; 65 ) ; sequencing problems ( 17 ; 67 ) ; highway problems ( 36 ; 63 )

Read the complete text