Balanced Contributions in Networks and Games with Externalities
Frank Huettner · 2026
Read Balanced Contributions in Networks and Games with Externalities by Frank Huettner (2026) 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-2604-13794.txt.
Introduction
Huettner characterizes the unique component-efficient rule for networks with externalities. Existence is the main challenge. The rule must satisfy balanced contributions on every edge. Construction uses only spanning-tree edges. Here, a cycle-sum identity bridges it, reducing non-tree contributions to subnetwork relations. The BCE rule matches the Myerson value for TU games and its Jackson-Wolinsky generalization without externalities. It restores the externality-free value. Unlike FCE, it never collapses to a graph-free formula. In short, the rule shares value fairly across the network.
Catalog record
| CREATOR | Frank Huettner |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 52666 |
| LICENSE | http://creativecommons.org/licenses/by-sa/4.0/ |
| RIGHTS | CC BY SA |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.13794
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Balanced Contributions in Networks and Games with Externalities Frank Huettner Abstract For networks with externalities, where each component’s worth may depend on the full network structure, balanced contributions and fairness lead to distinct component-efficient allocation rules. We characterize the unique component-efficient allocation rule satisfying balanced contributions — the BCE rule. Existence is the main challenge: balanced contributions must hold on every edge, but the construction uses only spanning-tree edges. A cycle-sum identity bridges this gap by reducing balanced contributions on non-tree edges to relations in proper subnetworks. The BCE rule coincides with the Myerson value for TU games and with its generalization by Jackson–Wolinsky for network games without externalities, it recovers the externality-free value on the complete network, and — unlike the fairness-based FCE rule — it does not reduce to a graph-free formula applied to the graph-restricted game. Keywords: networks with externalities; balanced contributions; component efficiency; Myerson value; worth function JEL Classification: C71, D62, D85 1 Introduction Myerson 1977a introduced a graph-based model of cooperation in which players are linked by a communication network, and only connected groups can cooperate effectively. For classical TU games, Myerson 1977a characterized an allocation rule by fairness and component efficiency; Myerson 1980 showed that component efficiency and balanced contributions yields the same value. Both axioms require connected players to benefit equally from their connection, but they impose different threats: fairness concerns removing a single link, while balanced contributions concerns a player’s complete withdrawal. 1 1 1 Osborne and Rubinstein 1994 , Ch. 14.4, derive balanced contributions from objections and counterobjections to proposed payoffs: player i objects by threatening to leave the game, and player j counters by noting that j ’s own departure would hurt i at least as much. The axiom requires that every such objection has a counterobjection. Jackson and Wolinsky 1996 enriched the model by allowing the worth of a component to depend on the full network structure, not just the coalition itself, and characterized the allocation rule satisfying component efficiency and fairness; Slikker 2007 showed that balanced contributions again yields the same rule. Navarro 2007 extended the model further, permitting externalities across components, and characterized the allocation rule satisfying fairness and component efficiency (henceforth FCE rule). With externalities, balanced contributions can capture indirect dependences that fairness misses, and leads to different payoffs. For example, consider the networks g and g^{\prime} in Figure 1 , and suppose player 3 gets a dollar whenever players 1 and 2 are linked; no other worth is generated. In g , player 3 is isolated and keeps the dollar by component efficiency. In g^{\prime} , the link \{1,3\} connects player 3 to player 1 , whose presence sustains the \{1,2\} -partnership: if player 1 withdraws, the partnership is destroyed, and player 3 ’s payoff vanishes. Balanced contributions accounts for this dependence and redistributes payoffs; fairness, which only considers removing the single link \{1,3\} when balancing players 1 and 3 , does not. g 1 2 3 g^{\prime} 1 2 3 Figure 1: Player 3 gets a dollar whenever players 1 and 2 are linked. In g , player 3 is isolated and keeps the dollar. In g^{\prime} , the link \{1,3\} exposes player 3 to player 1 ’s threats. Balanced contributions forces equal split (\tfrac{1}{3},\tfrac{1}{3},\tfrac{1}{3}) in g^{\prime} ; fairness leaves the payoff at (0,0,1) . However, an allocation rule satisfying balanced contributions and component efficiency in the presence of externalities has not been identified. The present paper fills this gap: we characterize the unique such rule (henceforth BCE rule), showing existence and uniqueness, and constructing it explicitly (Theorem 3 ). To this end, we impose balanced contributions only for directly connected players—a weaker requirement than pair-wise balanced contributions for all connected pairs, which we show to be incompatible with component efficiency. Moreover, it turns out that component efficiency, balanced contributions, and fairness are jointly incompatible in the presence of externalities. Whereas uniqueness follows from an argument going back to Myerson 1977a (within each component, one equation from component efficiency and |C|-1 equal gains equations along a spanning-tree pin down all |C| payoffs), existence is the actual challenge. Without externalities, existence is clear: the Shapley value of the graph-restricted game satisfies balanced contributions for all pairs of players, not just adjacent ones ( Slikker 2007 ) . With externalities, this pair-wise property fails, and the allocation rule defined via tree-edge balanced contributions equations must be shown to also satisfy balanced contributions on every non-tree edge. This is established through the cycle-sum identity (Lemma 4 ), which expresses balanced contributions on a non-tree edge as a signed sum of balanced contributions residuals in strict subnetworks, each zero by induction. Section 2 introduces the network game setting and the axioms. Section 3 constructs the BCE rule, states and proves the cycle-sum identity, and proves the main theorem. Section 4 investigates further properties and derives consequences. In particular, we recover the externality-free value ( Pham Do and Norde 2007 ; De Clippel and Serrano 2008 ) on the complete network without imposing a specific independence of externalities. Section 5 connects to games in partition function form (PFF), showing how the BCE and FCE characterizations extend to this framework, and that the BCE value—unlike the FCE value—does not reduce to a graph-free formula applied to the graph-restricted PFF game. 2 Preliminaries Let N=\{1,\ldots,n\} be a finite player set. A network (undirected graph) g on N is a set of unordered pairs \{i,j\} with i,j\in N , i\neq j , called links . Write g^{N}=\bigl\{\{i,j\}:i,j\in N,\,i\neq j\bigr\} for the complete network, and \mathcal{G} for the set of all networks on N . For S\subseteq N , write g|_{S}=\bigl\{\{k,l\}\in g:k\in S,\,l\in S\bigr\} for the subnetwork induced by S , and S/g for the partition of S into connected components of g|_{S} . In particular, N/g is the partition of N into connected components of g . For a partition \mathcal{P} of N , write g|_{\mathcal{P}}=\bigcup_{B\in\mathcal{P}}g|_{B} for the subnetwork of g keeping only links between players in the same block of \mathcal{P} . For a subset D\subseteq N , write g_{-D}\;=\;\bigl\{\{k,l\}\in g:k\notin D,\,l\notin D\bigr\} for the network with all links incident to any player in D deleted. For a single player j , write g_{-j}=g_{-\{j\}} . Since link removal is set-theoretic, g_{-i,-j}\;=\;g_{-j,-i}. A cycle of length \ell\geq 3 in g is a sequence of distinct players i_{0}\to i_{1}\to\cdots\to i_{\ell-1}\to i_{0} with each consecutive pair linked in g . A network without cycles is a forest ; a connected forest is a tree . We adopt the framework of Navarro 2007 . A worth function 2 2 2 Following Maschler et al. 2013 , we use worth rather than value to avoid confusion with allocation rules such as the Shapley value or the Myerson value. is a map w that assigns to each pair (C,g) , where C is a connected component of g , a real number w(C,g)\in\mathbb{R} . The worth w(C,g) is available to the members of component C when the network is g . Crucially, w(C,g) may depend on the full network g , not only on the internal structure g|_{C} ; this captures externalities across components. We write \mathcal{W} for the space of all worth functions. An allocation rule