An axiomatization of the Banzhaf value for cooperative games on antimatroids

Transcription

1 Math Meth Oper Res (24) 59:47 66 DOI.7/s86337 An axiomatization of the Banzhaf value for cooperative games on antimatroids E. Algaba, J. M. Bilbao 2, R. van den Brink 3, A. Jime nez-losada 4 Matemática Aplicada II, Escuela Superior de Ingenieros, Camino de los Descubrimientos, 492 Sevilla, Spain ( encarni@matinc.us.es) 2 Matemática Aplicada II ( mbilbao@us.es) 3 Department of Econometrics, Free University, De Boelelaan 5, 8 HV Amsterdam, The Netherlands ( jrbrink@feweb.vu.nl) 4 Matemática Aplicada II ( hispan@matinc.us.es) Manuscript received: January 23/Final version received: July 23 Abstract. Cooperative games on antimatroids are cooperative games in which coalition formation is restricted by a combinatorial structure which generalizes permission structures. These games group several well-known families of games which have important applications in economics and politics. The current paper establishes axioms that determine the restricted Banzhaf value for cooperative games on antimatroids. The set of given axioms generalizes the axiomatizations given for the Banzhaf permission values. We also give an axomatization of the restricted Banzhaf value for the smaller class of poset antimatroids. Finally, we apply the above results to auction situations. Key words: Antimatroid, Cooperative game, Permission structure, Banzhaf value Mathematics Subject Classification 2: 9A2 Introduction A cooperative game describes a situation in which a finite set of n players can generate certain payoffs by cooperation. A one-point solution concept for cooperative games is a function which assigns to every cooperative game a n-dimensional real vector which represents a payoff distribution over the players. The study of solution concepts is central in cooperative game theory. Two well-known solution concepts are the Shapley value as proposed by Shapley (953), and the Banzhaf value, initially introduced in the context of voting games by Banzhaf (965), and later on extended to arbitrary games by, e.g., Owen (975) and Dubey and Shapley (979). These values satisfy a set of * This research has been partially supported by the Spanish Ministery of Science and Technology, under grant SEC Financial support by the Netherlands Organization for Scientific Research (NWO), ESR-grant is gratefully acknowledged.

2 48 E. Algaba et al. intuitively reasonable axioms that characterizes each one of them. The assessment that both values assign to a player is the average of the marginal contribution to any coalition which the player belongs to, although they associate different weights to each coalition. The Banzhaf value considers that every player is equally likely to enter to any coalition whereas the Shapley value assumes that every player is equally likely to join to any coalition of the same size and all coalitions with the same size are equally likely. In a cooperative game the players are assumed to be socially identical in the sense that every player can cooperate with every other player. However, in practice there exist social asymmetries among the players. For this reason, the game theoretic analysis of decision processes in which one imposes asymmetric constraints on the behavior of the players has been and continues to be an important subject to study. Important consequences have been obtained of adopting this type of restrictions on economic behavior. Some models which analyze social asymmetries among players in a cooperative game are described in, e.g., Myerson (977), Owen (986) and Borm, Owen and Tijs (992). In these models the possibilities of coalition formation are determined by the positions of the players in a communication graph. Another type of asymmetry among the players in a cooperative game is introduced in Gilles, Owen and van den Brink (992), Gilles and Owen (999), van den Brink and Gilles (996) and van den Brink (997). In these models, the possibilities of coalition formation are determined by the positions of the players in a hierarchical permission structure. Two different approaches were introduced for these games: conjunctive and disjunctive. Games on antimatroids were introduced in Jime nez-losada (998). Later, Algaba, Bilbao, van den Brink and Jime nez-losada (2) showed that the feasible coalition systems derived from both the conjunctive and disjunctive approach to games with a permission structure were identified to certain families of antimatroids: poset antimatroids and antimatroids with the path property, respectively. On the onther hand, Branzei, Fragnelli and Tijs (22) introduced peer group games that were described by a rooted tree. This type of games allows to study particular cases of auction situations, communication situations, sequencing situations and flow games. These games are restricted games on poset antimatroids wiht the path property. This class of antimatroids are the permission forest and permission tree structures which are often encountred in the economic literature. So, the study of games on antimatroids allows to unify several research lines. Another model in which cooperation possibilities in a game are limited by some hierarchical structure on the set of players can be found in Faigle and Kern (992) who consider feasible rankings of the players. A relevant aspect in permission structures has been the study and characterization of solution concepts, the so-called permission values, defined with the aid of the Shapley value of two different permission games (see van den Brink and Gilles (996) and van den Brink (997)) and the permission Banzhaf values based on the Banzhaf value for these games (see van den Brink (2)). However, as we have already pointed out antimatroids extend this model allowing us to study new situations and at the same time unify different approaches. One of the important aspects of paying attention to cooperative games on antimatroids is revealed in the study of a characterization of the Shapley value for cooperative games on antimatroids in Algaba, Bilbao, van den Brink and Jime nez-losada (23) where in particular we unify the

3 An axiomatization of the Banzhaf value for cooperative games on antimatroids 49 fairness axioms used in both the conjunctive and disjunctive approaches. This allows us at the same time to simplify the existing literature as well as provide new insights. The current paper complements this study by focusing on the Banzhaf value obtaining an axiomatization for Banzhaf value by means of intuitive reasonable axioms and giving answers to the questions that were studied for the Shapley value. In the last section, we emphasize by means of an example both the importance of the axiomatization as well as the differences of interpretation between the two values studied. The paper is organized as follows. To make the paper self-contained Section 2 recalls some preliminaries on cooperative games, antimatroids and permission structures. Section 3 contains the main result which is focussed on an axiomatization of the restricted Banzhaf value for games on antimatroids. The set of given axioms generalizes the axiomatizations of both the conjunctive and disjunctive Banzhaf permission values for games with a permission structure. In particular, with respect to these we unify the fairness axioms as well as the predecessor fairness used in both conjunctive and disjunctive approaches to a stronger fairness axiom. In Section 4, we restrict our attention to a special class of antimatroids the so-called poset antimatroids, showing that deleting the strong fairness axiom characterizes the restricted Banzhaf value for the class of cooperative games on poset antimatroids. Moreover, it turns out that the class of games on poset antimatroids is characterized as that class of games on which the restricted Banzhaf value is the unique solution satisfying these axioms. Finally, we apply the obtained results to auction situations. In particular, we obtain an axiomatization of the Banzhaf value in this context as well as a nice formula to compute it. 2 Cooperative games on antimatroids A cooperative game is a pair ðn; vþ, where N N is a finite set of players and v : 2 N! R is a characteristic function satisfying vð;þ ¼. A cooperative game ðn; vþ is monotone if vðeþ vðf Þ whenever E F N. Antimatroids were introduced by Dilworth (94) as particular examples of semimodular lattices. A symmetric study of these structures was started by Edelman and Jamison (985) emphasizing the combinatorial abstraction of convexity. The convex geometries are a dual concept of antimatroids (see Bilbao, 2). An antimatroid A on N is a family of subsets of 2 N, satisfying A. ;2A. A2. (Accessibility) If E 2A, E 6¼ ;, then there exists i 2 E such that Enfig 2A: A3. (Closed under union) If E; F 2Athen E F 2A. The definition of antimatroid implies the following augmentation property: if E; F 2Awith jej > jf j then there exists i 2 EnF such that F fig 2A. From now on, we only consider antimatroids satisfying A4. (Normality) For every i 2 N there exists an E 2Asuch that i 2 E. In particular, this implies that N 2A. Now we introduce some well-known concepts about antimatroids which can be found in Korte, Lova sz and

4 5 E. Algaba et al. Schrader (99, Chapter III). Let A be an antimatroid on N. This set family allows S to define the interior operator int A : 2 N!A, given by int A ðeþ ¼ F E;F 2A F 2A, for all E N. Let A be an antimatroid on N. Anaugmentation point (Jiménez-Losada, 998) of E 2Ais a player i 2 NnE such that E fig 2A, i.e., those players that can be joined to a feasible coalition keeping feasibility. In a dual way, an extreme point (Edelman and Jamison, 985) of E 2Ais a player i 2 E such that Enfig 2A, i.e., those players that can leave a feasible coalition E keeping feasibility. By condition A2 (Accessibility) every non-empty coalition in A has at least one extreme point. A set E 2Ais a path in A if it has a single extreme point. The path E 2Ais called a i-path in A if it has i 2 N as unique extreme point. A coalition E 2Aif and only if E is a union of paths. Moreover, for every E 2Awith i 2 E there exists an i-path F such that F E. The set of i-paths for a given player i 2 N will be denoted by AðiÞ. A special class of antimatroids are the poset antimatroids being antimatroids that are closed under intersection. An antimatroid A is a poset antimatroid if E \ F 2A for every E; F 2A. Alternatively, the poset antimatroids are characterized as those antimatroids such that for every i 2 N there is exactly one i-path. For a cooperative game ðn; vþ and an antimatroid A on N we define the restricted characteristic function v A which assigns to every coalition E the worth generated by the interior of E, i.e., v A ðeþ ¼vðint A ðeþþ, for all E N: A solution for games on antimatroids is a function f that assigns a payoff distribution f ðn; v; AÞ 2 R n to every cooperative game ðn; vþ and antimatroid A on N. Algaba et al. (23) characterized the restricted Shapley value for cooperative games on antimatroids. In this paper we characterize the restricted Banzhaf value BðN; v; AÞ for a cooperative game ðn; vþ and an antimatroid A on N which is obtained by applying the Banzhaf value to game ðn; v A Þ, i.e., B i ðn; v; AÞ ¼ B i ðn; v A Þ¼ X ðv 2 jnj A ðeþ v A ðenfigþþ: fen:i2eg The Banzhaf value was introduced as a power index for voting games by Banzhaf (965), and later was generalized to arbitrary games by, e.g., Owen (975) and Dubey and Shapley (979). As we have already indicated games on antimatroids generalize cooperative games with an acyclic permission structure. In the conjunctive approach as developed in Gilles et al. (992), it is assumed that each player needs permission from all its predecessors before it is allowed to cooperate. Alternatively, in the disjunctive approach as discussed in Gilles and Owen (999) it is assumed that each player that has predecessors only needs permission from at least one of its predecessors before it is allowed to cooperate with other players. Algaba et al. (2) show that for every acyclic permission structure, both the feasible coalition set derived from the conjunctive and disjunctive approaches are antimatroids. Moreover, the class of all sets of feasible coalitions that can be obtained as conjunctive feasible coalitions is exactly the class of poset antimatroids. The class of all sets of feasible coalitions that can be obtained as disjunctive feasible coalitions is exactly the class of antimatroids satisfying the so-called path property. A solution for games with a permission structure is a function f that assigns a payoff distribution f ðn; v; SÞ 2R n to every cooperative game

5 An axiomatization of the Banzhaf value for cooperative games on antimatroids 5 ðn; vþ and permission structure S on N. The conjunctive Banzhaf permission value is obtained by applying the Banzhaf value to the conjunctive restricted games, while the disjunctive Banzhaf permission value is obtained by applying the Banzhaf value to the disjunctive restricted games. (See van den Brink (2)). The aim in the next sections will be to generalize axiomatizations given for the conjunctive and disjunctive Banzhaf permission values to obtain an axiomatization of the restricted Banzhaf value for cooperative games on antimatroids. 3 An axiomatization of the restricted Banzhaf value In this section an axiomatization of the restricted Banzhaf value for games on antimatroids is provided. This result generalizes the axiomatizations of the conjunctive and disjunctive Banzhaf permission values given in van den Brink (2). The first axiom is a weaker version of the usual efficiency axiom. Axiom (One-player efficiency). For every cooperative game ðn; vþ and antimatroid A on N, if N ¼fig then f i ðn; v; AÞ ¼ vðfigþ. The next three axioms are generalizations of corresponding axioms for cooperative games with a permission structure and are already stated in Algaba et al. (23). For two cooperative games ðn; vþ and ðn; wþ the game ðn; v þ wþ is given by ðv þ wþðeþ ¼vðEÞþwðEÞ for all E N. Axiom 2 (Additivity). For every pair of cooperative games ðn; vþ; ðn; wþ and antimatroid A on N, f ðn; v þ w; AÞ ¼ f ðn; v; AÞ þ f ðn; w; AÞ. Axiom 3 (Necessary player property). For every monotone cooperative game ðn; vþ and antimatroid A on N, if i 2 N satisfies vðeþ ¼ for all E Nnfig then f i ðn; v; AÞ f j ðn; v; AÞ for all j 2 N. Given the antimatroid A on N, the basic path group P i of player i is given by those players that are in every i-path, i.e., P i ¼ T E2AðiÞ E. This set is formed by those players that totally control player i in A, i.e., without them player i can not be in any feasible coalition. Obviously, i 2 P i for all i 2 N. Axiom 4 (Structural monotonicity). For every monotone cooperative game ðn; vþ and antimatroid A on N, if j 2 N then for all i 2 P j we have f i ðn; v; AÞ f j ðn; v; AÞ. Before introducing three new axioms for cooperative games on antimatroids, we introduce some concepts. For cooperative game ðn; vþ and j 2 N define cooperative game ðnnfjg; v j Þ by v j ðeþ ¼vðEÞ for all E Nnfjg. For antimatroid A on N define the set system A j on Nnfjg by A j ¼fEnfjg: E 2Ag: So, A j is obtained by deleting j from all coalitions in A. Next we establish that A j is an antimatroid. Lemma. If A is an antimatroid on N and j 2 N then A j is an antimatroid on Nnfjg.

6 52 E. Algaba et al. Proof. (i) As ;2Awe have that ;2A j. (ii) Suppose that E; F 2A j. Then we can establish the following: (a) if E; F 2A then E F 2A, and (since j =2 E F ), E F 2A j ; (b) if fe; F g 6 A then E F fjg 2A, and thus E F 2A j. So, A j is closed under union. (iii) Consider E 2A j. We establish the following: (a) if E 2A then there exists i 2 E such that Enfig 2A. Since j =2 E, it holds that Enfig 2A j ; (b) if E =2 Athen E fjg 2A, implying that there exists i 2 E (clearly i 6¼ j) such that ðe fjgþnfig 2A. But then Enfig 2A j. So, A j satisfies accessibility. (iv) Finally, it follows straightforward that A j satisfies normality, i.e., Nnfjg 2A j. h Given the antimatroid A on N, the path group P i of player i is defined as the set of players that are in some i-path, i.e., P i ¼ S E2AðiÞ E. So, the path group of player i are all those players of which i has some dependence, i.e., these players in some sense partially control player i. Obviously, P i P i for all i 2 N. Now, given an antimatroid A on N, we call i 2 N an inessential player for A in ðn; vþ if player i and every player j 2 N such that i 2 P j are null players in ðn; vþ. The first of the three new axioms is related to the null player out property of the Banzhaf value (see Derks and Haller, 994) and states in addition to the inessential player property as introduced in Algaba et al. (23), that deleting an inessential player from a game on an antimatroid does not change the payoffs of the other players. Axiom 5 (Strong inessential player property). For every cooperative game ðn; vþ and antimatroid A on N, if i is an inessential player for A in ðn; vþ then if j ¼ i, f j ðn; v; AÞ ¼ f j ðnnfig; v i ; A i Þ otherwise. As we have already pointed out the fairness axiom introduced in Algaba et al. (23) generalizes both conjunctive and disjunctive fairness for games with a permission structure (see van den Brink 997, 999). In order to provide a characterization of the restricted Banzhaf value we need to strengthen this axiom in a way such that it also generalizes predecessor fairness for games with a permission structure (see van den Brink, 2). For antimatroid A on N and coalition E 2A, let au A ðeþ be the set of all augmentation points of E in A. The new strong fairness axiom states that deleting the feasible coalition E from the set of feasible coalitions (as long as AnfEg is still an antimatroid) changes the payoffs of all players in E by the same amount and, moreover, the payoffs of the augmentation points of E in A also change by this same amount but in opposite direction with respect to the elements in E. Axiom 6 (Strong fairness). For every cooperative game ðn; vþ, antimatroid A on N, and E 2Awith jej 2 such that AnfEg is an antimatroid on N, it holds that for all i 2 E and j 2 au A ðeþ f i ðn; v; AÞ f i ðn; v; AnfEgÞ ¼ f j ðn; v; AnfEgÞ f j ðn; v; AÞ:

7 An axiomatization of the Banzhaf value for cooperative games on antimatroids 53 The following result (given in Algaba et al. (23)) establishes under what conditions a coalition E can be deleted, AnfEg being still an antimatroid. Lemma 2. Let A be an antimatroid on N and E 2A. Then, AnfEg is an antimatroid on N if and only if E is a path, E =2 f;; N g and every F 2A satisfying E F and jf j¼jejþ is not a path. The last axiom is based on the proxy neutrality and amalgamation properties of the Banzhaf value as considered in Lehrer (988) and Haller (994). For antimatroid A on N and player j 2 N define A j on N by A j ¼fE 2A: j 2 Eg, i.e., A j is the set of coalitions in A that contain player j. Further, for player h 2 NnN we define A hj by A hj ¼fE fhg: E 2A j gfenfjgfhg: E 2A j g ðana j Þ: Note that AnA j need not be equal to A j. In some sense player h 2 NnN is controlling the player j when going from antimatroid A to A hj. Feasible coalitions containing player j now should also contain player h. Moreover, all feasible coalitions in A that contain player j are still feasible if we replace player j by player h. Finally, all feasible coalitions in A that do not contain player j stay feasible. We first establish that A hj is an antimatroid. Lemma 3. If A is an antimatroid on N, then for every j 2 N and h 2 NnN, A hj is an antimatroid on N fhg. Proof. (i) If ;2Athen ; 2 AnA j, and thus ;2A hj. (ii) For E 2A hj we establish that: (a) if h =2 then j =2 E, and thus E 2 AnA j ; (b) if h 2 E and j =2 E then Enfhgfjg 2A; (c) if h 2 E and j 2 E then Enfhgfjg ¼Enfhg 2A. Suppose that E; F 2A hj. With the cases (a), (b) and (c) it follows that: () if h =2F then E; F 2 AnA j, and thus E F 2 AnA j A hj ; (2) if h 2 E F then ½Enfhgfjg; F nfhgfjg 2AŠ or ½Enfhg 2 AnA j and F nfhgfjg 2AŠ or ½Enfhgfjg 2A and F nfhg 2 AnA j Š. In all these cases ðe F fjgþnfhg 2A. But then E F 2A hj. So, A hj is closed under union. (iii) For E 2A hj we have that: (a) if j 2 E then h 2 E, and thus Enfjg 2A hj ; (b) if j =2 E and h =2 E then E 2 AnA j ; and thus there exists i 2 E such that Enfig 2 AnA j A hj ; (c) if j =2 E and h 2 E then E fjgnfhg 2A and thus there exists k 2 E fjgnfhg such that E fjgnfh; kg 2A.Ifk¼j then E fjgnfh; kg ¼Enfhg 2 AnA j A hj. Otherwise k 6¼ j and E fjgnfh; kg 2A, implying that Enfkg 2A hj. So, A hj satisfies accessibility. (iv) Finally, it follows straightforward that A hj satisfies normality, i.e., N fhg 2A hj : h Besides letting player h control player j in the antimatroid we also let h control j in the game by requiring the presence of h for any non-zero contribution of j. This is the only contribution of player h. This means that we consider the game ðn fhg; v hj Þ with v hj on N fhg given by v hj ðeþ ¼vðEnfjgÞ if h =2 E, andv hj ðeþ ¼vðEnfhgÞ otherwise. The next axiom is required for a special class of antimatroids. Note that an antimatroid always has at least jnjþ elements. If jaj ¼ jnjþ then there is

8 54 E. Algaba et al. a unique coalition in A of cardinality i from i ¼ until i ¼ n. So, there exists a unique i-path for every player i. We refer to such an antimatroid as a chain antimatroid. For chain antimatroids we require that the sum of payoffs of players h 2 NnN and j 2 N after letting player h control player j is equal to the payoff of player j when h is not yet present. Axiom 7 (Proxy neutrality). For every cooperative game ðn; vþ, chain antimatroid A on N, j 2 N and h 2 NnN, it holds that f h ðn fhg; v hj ; A hj Þþf j ðn fhg; v hj ; A hj Þ¼f j ðn; v; AÞ: Applying this axiom to games with a permission structure yields an axiom that is weaker than vertical neutrality and neither weaker nor stronger than horizontal neutrality for games with a permission structure (see van den Brink, 2). A stronger version of proxy neutrality, namely the one which states that f h ðn fhg; v hj ; A hj Þþf j ðn fhg; v hj ; A hj Þ¼f j ðn; v; AÞ for any antimatroid A on N and players j 2 N and h 2 NnN, generalizes both vertical and horizontal neutrality for games with a permission structure. Although this stronger version is satisfied by the restricted Banzhaf value for games on antimatroids, for the characterization result presented in Theorem 2 it is sufficient to require this axiom just for chain antimatroids. Since we use the stronger version in our application in Section 5, we also formally state this version and prove that the restricted Banzhaf value satisfies this axiom. Axiom 8 (Strong proxy neutrality). For every cooperative game ðn; vþ, antimatroid A on N, j 2 N and h 2 NnN, it holds that f h ðn fhg; v hj ; A hj Þþf j ðn fhg; v hj ; A hj Þ¼f j ðn; v; AÞ: The next result yields that the restricted Banzhaf value satisfies all axioms discussed above. Theorem. The restricted Banzhaf value B for cooperative games on antimatroids satisfies one-player efficiency, additivity, the necessary player property, structural monotonicity, the strong inessential player property, strong fairness and strong proxy neutrality. Proof. Let ðn; vþ be a cooperative game and A be an antimatroid on N.. If N ¼fig then A¼f;; Ng, and thus v ¼ v A. One-player efficiency then follows from one-player efficiency of the Banzhaf value for cooperative games. 2, 3 and 4. Additivity, the necessary player property and structural monotonicity of the restricted Banzhaf value follow in a similar way as for the restricted Shapley value in Algaba, Bilbao, van den Brink and Jime nez- Losada (23). (To show the necessary player property and structural monotonicity take Banzhaf weights p e ¼ instead of Shapley 2 jnj weights pe ¼ ðjnj jejþ!ðjej Þ! jnj! ). 5. If i is an inessential player for A in ðn; vþ then i is a null player in ðn; v A Þ. The null player property of the Banzhaf value then implies that f i ðn; v; AÞ ¼. From the null player out property of the Banzhaf value

9 An axiomatization of the Banzhaf value for cooperative games on antimatroids 55 (see Derks and Haller, 994) it follows that f j ðn; v; AÞ ¼ f j ðnnfig; v i ; A i Þ for j 6¼ i. 6. Let E 2Abe such that AnfEg is an antimatroid on N. For i 2 E: (a) If E 6 F then int A ðf Þ¼ fh2a:hf g H ¼ fh2anfeg:hf g H ¼ int AnfEg ðf Þ: Hence, v A ðf Þ¼v AnfEg ðf Þ. In particular, as i 2 E it holds that E 6 F nfig and therefore v A ðf nfigþ ¼ v AnfEg ðf nfigþ. (b) If there exists j 2 au A ðeþ such that E fjg F then int A ðf Þ¼ int AnfEg ðf Þ. Hence also in this case, v A ðf Þ¼v AnfEg ðf Þ. Since i 2 E and E F implies that i 2 F, it follows for i 2 E and j 2 au A ðeþ that B i ðn; v; AÞ B i ðn; v; AnfEgÞ ¼ B i ðn; v A Þ B i ðn; v AnfEg Þ ¼ X v 2 jnj A ðf Þ v A ðf nfigþ v AnfEg ðf Þþv AnfEg ðf nfigþ ff N:i2F g ¼ 2 jnj X ff N:EF ;F \au A ðeþ¼;g ¼ X 2 jnj ff N:EF ;F \au A ðeþ¼fjgg ¼ X 2 jnj ff N:j2F g ¼ B j N; v AnfEg Bj ðn; v A Þ ¼ B j ðn; v; AnfEgÞ B j ðn; v; AÞ; showing that B satisfies strong fairness. v A ðf Þ v AnfEg ðf Þ v A ðf nfjgþ v AnfEg ðf nfjgþ v A ðf nfjgþ v AnfEg ðf nfjgþ þ v AnfEg ðf Þ v A ðf Þ 7. Let j 2 N and h 2 NnN. We establish the following facts: (a) if fh; jg 6 F then v hj ðhþ ¼v hj ðhnfh; jgþ ¼ vðhnfh; jgþ for all H F. (b) since F 2AnA j implies that h =2 F, and F A hj na j implies that h 2 F, we have for F N fhg and h 2 F that ðv hj Þ Ahj ðf Þ¼v hj int Ahj ðf Þ ¼ HA ¼ v hj H2fEfhgjE2A j g HF H fh2a hj :HF g H2fEnfjgfhgjE2A j g HF H H2AnA j HF HC A

10 56 E. Algaba et al. ¼ v ¼ v@ ¼ v@ fh2a j :HF g fh2a j :HF nfhgg fh2a:hf nfhgg ðh fhgþ H HA fh2ana j :HF g fh2ana j :HF nfhgg ¼ vðint A ðf nfhgþþ ¼ v A ðf nfhgþ; HA HA and ðv hj ÞA hj ðf nfhgþ ¼ v hj int Ahj ðf nfhgþ ¼ ¼ v hj H2fEfhgjE2A j g HF nfhg ¼ v ¼ v@ H fh2ana j :HF nfhgg fh2a:hf nfh;jgg H2fEnfjgfhgjE2A j g HF nfhg HA ¼ v@ HA ¼ vðint A ðf nfh; jgþþ ¼ v A ðf nfh; jgþ: fh2a hj :HF nfhgg H fh2ana j :HF nfhgg H2AnA j HF nfhg HA HA HC A With this it follows that B h ðn fhg; v hj ; A hj Þþ B j ðn fhg; v hj ; A hj Þ ¼ B h ðn fhg; ðv hj Þ Ahj ÞþB j ðn fhg; ðv hj Þ Ahj Þ ¼ X ðv 2 jnj hj Þ Ahj ðf Þ ðv hj Þ Ahj ðf nfhgþ ff Nfhg:h2F g þ 2 jnj X ff Nfhg:j2F g ¼ 2 jnj X ff Nfhg:fh;jgF g þ 2 jnj X ff Nfhg:fh;jgF g ðv hj Þ Ahj ðf Þ ðv hj Þ Ahj ðf nfjgþ ðv hj Þ Ahj ðf Þ ðv hj Þ Ahj ðf nfhgþ ðv hj Þ Ahj ðf Þ ðv hj Þ Ahj ðf nfjgþ

11 An axiomatization of the Banzhaf value for cooperative games on antimatroids 57 ¼ 2 jnj X ff Nfhg:fh;jgF g ðv A ðf nfhgþ v A ðf nfh;jgþþ þ X ðv 2 jnj A ðf nfhgþ v A ðf nfh;jgþþ ff Nfhg:fh;jgF g ¼ 2 jnj X ff N:j2F g ¼ 2 jnj X ff N:j2F g ðv A ðf Þ v A ðf nfjgþþþ X ff N:j2F g ðv A ðf Þ v A ðf nfjgþþ ððv A ðf Þ v A ðf nfjgþþ¼b j ðn;v A Þ¼ B j ðn;v;aþ; showing that B satisfies strong proxy neutrality. h To prove uniqueness we first prove the following lemma. Lemma 4. Let A be an antimatroid on N and let i; j 2 N. Ifi 2 P j np j then there is an E 2Awith i; j 2 E such that AnfEg is an antimatroid. Proof. Let i 2 P j np j, then j has at least two paths one of them contain i and the other path does not contain i. Let E 2 AðjÞ, E 6¼ N be a j-path containing i. By Lemma 2 one of the following two cases happen: AnfEg is an antimatroid or there exists k 2 NnE with E fk g a k -path. If AnfE fk gg is an antimatroid the proof is finished, otherwise applying again Lemma 2 there exists k 2 with E fk ; k 2 g a k 2 -path. Repeating this reasoning and assuming that in each step the path obtained is not removable then we obtain a chain of paths from E to N (with at least one coalition different from E). By proving that this case is not possible, the lemma will be shown. Let F be a j-path such that i =2 F and let k be the last element in F in incorporating to E in the chain above and H the path in this step of the chain. We have H 6¼ E is a k-path, E F H and k 2 F : This is a contradiction because there exists a k-path H contained in F and therefore in some other k-path, since H contains i and H does not. h The axioms discussed above characterize the restricted Banzhaf value for cooperative games on antimatroids. Theorem 2. A solution f for cooperative games on antimatroids is equal to the restricted Banzhaf value B if and only if it satisfies one-player efficiency, additivity, the necessary player property, structural monotonicity, the strong inessential player property, strong fairness and proxy neutrality. Proof. To prove uniqueness, suppose that solution f satisfies the seven axioms. Consider antimatroid A on N and the monotone game w T ¼ c T u T, c T, where u T is the unanimity game of T N, i.e., w T ðeþ ¼c T if E T, and w T ðeþ ¼ otherwise. If c T ¼ then the strong inessential player property implies that f i ðn; w T ; AÞ ¼ for all i 2 N. Suppose that c T >. We show that f ðn; w T ; AÞ is uniquely determined by induction on jnj. If N ¼fig then f ðn; w T ; AÞ is uniquely determined by one-player efficiency. Proceeding by induction assume that f ðn ; w T ; AÞ is uniquely determined if jn j < jnj. We use a second induction step on jaj. (Note that jaj jn j þ ). Consider the sets P T ¼ S i2t P i and P T ¼ S i2t P i.!

12 58 E. Algaba et al.. If jaj ¼ jnjþ then there is a unique coalition in A of cardinality i from i ¼ until i ¼ n. So, there exists a unique i-path for every player i. In this case, we have that P T ¼ P T. We distinguish the following two cases: (i) Suppose that P T 6¼ N. Then there exist inessential players. Take an inessential player j. The strong inessential player property implies that f j ðn; w T ; AÞ ¼ and f i ðn; w T ; AÞ ¼ f i ðnnfjg; ðw T Þ j ; A j Þ for every i 2 Nnfjg. By the induction hypothesis f i ðn; w T ; AÞ is uniquely determined for all i 2 Nnfjg. (ii) Suppose that P T ¼ N. Then there exist i ; i 2 N such that P i ¼ N and P i ¼ P i nfi g. Since ða i Þ i i ¼A and w T nfii g i i ¼ w T, proxy neutrality implies that f i ðn; w T ; AÞ þ f i ðn; w T ; AÞ ¼ f i Nnfi g; w T nfi g; A i : ðþ Structural monotonicity and the necessary player property imply that there exists c 2 R such that f i ðn; w T ; AÞ ¼ c for all i 2 P T ¼ N. With () it follows that c ¼ 2 f i Nnfi g; w T nfi g; A i. With the induction hypothesis c is uniquely determined, and so is fðn; w T ; AÞ. 2. Now we suppose that jaj > jnjþ. Proceeding by induction assume that f N; w T ; A is uniquely determined if A < jaj. Again we distinguish the following two cases: (i) Suppose that P T 6¼ N. Then fðn; w T ; AÞ is uniquely determined by the strong inessential player property and the induction hypothesis similarly as shown above for the case jaj ¼jNjþ. (ii) Suppose that P T ¼ N. Ifi 2 P T then structural monotonicity and the necessary player property imply that there exists c 2 R such that f i ðn; w T ; AÞ ¼ c. If i 2 P T np T then there exists j 2 T with i 2 P j np j. By Lemma 4 there exists an E 2Awith i; j 2 E such that AnfEg is an antimatroid. Applying strong fairness then yields that f i ðn; w T ; AÞ ¼ c c i with c i ¼ f j N; w T ; A fi N; w T ; A for an A with A ¼ jaj. On the other hand, as E 6¼ N there exists k 2 au A ðeþ and thus k =2 E. Applying strong fairness, it holds for i 2 E that f i ðn; w T ; AÞ f i N; w T ; A ¼ fk N; w T ; A fk ðn; w T ; AÞ; where A ¼AnfEg. Since f k ðn; w T ; AÞ ¼ c c k, denoting by c k ¼ f i N; w T ; A þ fk N; w T ; A it follows that c c i f i ðn; w T ; A Þ¼f k ðn; w T ; A Þ cþc k implying that c ¼ 2 ð c k þ c k þ c i Þ (like c i, c k and c k are determined by the induction hypothesis). With cases (i) and (ii) above f i ðn; w T ; AÞis determined for all i 2 N. Above we showed that fðn; w T ; AÞ is uniquely determined for all (monotone) games w T ¼ c T u T with c T. Suppose that w T ¼ c T u T with c T <. (Then w T is not monotone and the necessary player property and

13 An axiomatization of the Banzhaf value for cooperative games on antimatroids 59 structural monotonicity cannot be applied.) Let v 2G N denote the null game, i.e., v ðeþ ¼ for all E N. From the strong inessential player property it follows that f i ðn; v ; A Þ ¼ for all i 2 N. Since w T ¼ c T u T with c T, and ðv Þ ¼ ðw T Þ A þ ðw T Þ A, it follows from additivity of f that fðn; w T ; AÞ ¼ fðn; v ; AÞ fðn; w T ; AÞ ¼ fðn; w T ; AÞ is uniquely determined because w T is monotone. So, fðn; c T u T ; AÞ is uniquely determined for all c T 2 R. Since every cooperative game v on N can be expressed as a linear combination of unanimity games it follows with additivity that fðn; v; AÞ is uniquely determined. h We end this section by showing logical independence of the axioms stated in Theorem.. The zero solution given by f i ðn; v; AÞ ¼ for all i 2 N satisfies additivity, the necessary player property, structural monotonicity, the strong inessential player property, strong fairness and proxy neutrality. It does not satisfy one-player efficiency. 2. Let DðvÞ ¼maxfd v ðt Þ : T Ng, where d v ðt Þ¼ X F Tð Þ jt j jf j vðf Þ is the Harsanyi dividend for T and DðvÞ ¼fT N : d v ðt Þ¼DðvÞg. The solution given by fðn; v; AÞ ¼ B N; P T 2DðvÞ d vðt Þu T ; A satisfies oneplayer efficiency, the necessary player property, structural monotonicity, the strong inessential player property, strong fairness and proxy neutrality. It does not satisfy additivity. 3. Let f i ðn; v; AÞ ¼ vn ð Þ vnnfig ð Þ if fig 2Aand f i ðn; v; AÞ ¼ otherwise. This solution satisfies one-player efficiency, additivity, structural monotonicity, the strong inessential player property, strong fairness and proxy neutrality. It does not satisfy the necessary player property. 4. The solution given by fðn; v; AÞ¼BðN; vþ satisfies one-player efficiency, additivity, the necessary player property, the strong inessential player property, strong fairness and proxy neutrality. It does not satisfy structural monotonicity. 5. The solution f i ðn; v; A Þ ¼ vðnþ 2 jnj for all i 2 N, satisfies one-player efficiency, additivity, the necessary player property, structural monotonicity, strong fairness and proxy neutrality. It does not satisfy the strong inessential player property. 6. For antimatroid A on N and player i 2 N, consider P i ¼ T E2AðiÞ E.To every antimatroid A on N we associate the poset antimatroid A P that is obtained by taking all coalitions that can be obtained as unions of coalitions in fp i : i 2 Ng. The solution f i ðn; v; AÞ ¼ BðN; v; A P Þ satisfies oneplayer efficiency, additivity, the necessary player property, structural monotonicity, the strong inessential player property and proxy neutrality. It does not satisfy strong fairness.

14 6 E. Algaba et al. 7. The solution f i ðn; v; AÞ ¼ vðfigþ if fig 2Aand f i ðn; v; AÞ ¼ otherwise satisfies one-player efficiency, additivity, the necessary player property, structural monotonicity, the strong inessential player property and strong fairness. It does not satisfy proxy neutrality. 4 Poset antimatroids Deleting strong fairness from the set of axioms stated in Theorem the remaining set of axioms characterizes the restricted Banzhaf value for games on poset antimatroids. Moreover, poset antimatroids are the unique antimatroids for which it is possible to delete the strong fairness axiom. In order to state this result we first have to show that the antimatroids A j and A hj, where j 2 N; h 2 NnN are poset antimatroids whenever A is a poset antimatroid. Lemma 5 Let A be a poset antimatroid on N. If j 2 N and h 2 NnN then (i) A j is a poset antimatroid on Nnfjg, and (ii) A hj is a poset antimatroid on N fhg. Proof. From Lemma s and 3 it already follows that A j, and A hj are antimatroids. So, we only have to show that they are closed under intersection. (i) Suppose that E; F 2A j. (a) if E; F 2A then E \ F 2A, and (since j =2 E \ F ), E \ F 2A j. (b) If fe; F g 6 Athen assume without loss of generality that E fjg 2A. If also F fjg 2A then ðe \ F Þfjg ¼ ðe fjgþ \ ðf fjgþ 2 A, and thus E \ F 2A j. Otherwise, if F fjg =2 A then j =2 F and F 2A, which implies that j =2 E \ F ¼ ðe fjgþ \ F 2A, and thus E \ F 2A j. So, A j is closed under intersection. (ii) Suppose that E; F 2A hj. (a) If h 2 E \ F then Enfhgfjg; F nfhgfjg 2 A, and thus E \½FnfhgfjgŠ ¼ ½EnfhgfjgŠ \ ½F nfhgfjgš 2 A. But then E \ F 2A hj. (b) If h =2 E \ F then assume without loss of generality that h =2 E. Then E 2AnA j. If also h =2 F then E; F 2 AnA j, and thus E \ F 2AnA j A hj. Otherwise, if h 2 F then F nfhgfjg 2A. But then j =2 E \ F ¼ E \ðfnfhgþ ¼ E \½FnfhgfjgŠ 2 A. But then E \ F 2A hj. So, A hj is closed under intersection. h Note that poset antimatroids are the unique antimatroids such that every player has a unique path. In particular, we can conclude that given an antimatroid A on N, then A is a poset antimatroid if and only if P i ¼ P i, for all i 2 N. Theorem 3. A solution f for games on poset antimatroids is equal to the restricted Banzhaf value if and only if it satisfies one-player efficiency, additivity, the necessary player property, structural monotonicity, the strong inessential player property and proxy neutrality. Proof. Suppose that solution f satisfies the six axioms on poset antimatroids. Consider a poset antimatroid A on N and the game w T ¼ c T u T, c T. Taking into account that for a poset antimatroid P T ¼ P T the proof follows of the first part from Theorem 2. For arbitrary v it follows that fðn; v; AÞ is uniquely determined in a similar way as in the proof of Theorem 2. h

15 An axiomatization of the Banzhaf value for cooperative games on antimatroids 6 Hence, on poset antimatroids the restricted Banzhaf value is uniquely determined without using strong fairness. The solutions, 2, 3, 4, 5 and 7 given at the end of the previous section show logical independence of the axioms stated in Theorem 3. In Algaba et al. (2) it is shown that for every i 2 N, B i ðn; v; AÞ ¼ X d va ðeþ E2A i 2 : jej Next, we remark that for games on poset antimatroids the restricted Banzhaf value can be written using dividends of the original game as follows. Proposition. If A is a poset antimatroid on N then B i ðn; v; AÞ ¼ X d v ðt Þ 2 : jp T j ft N:i2P T g Proof. Since d v ðeþ ¼ for every E =2 Ait follows that B i ðn; v; AÞ ¼ X d va ðeþ 2 ¼ X P ft E:E¼P T g d vðt Þ jej 2 jej E2A i E2A i ¼ X d v ðt Þ 2 : h jp T j ft N:i2P T g We can characterize the class of cooperative games on poset antimatroids as the class of games on which the restricted Banzhaf value satisfies the six axioms of Theorem 3. Given an antimatroid A on N, let B ð; AÞ be the function that assigns to every cooperative game ðn; vþ the restricted Banzhaf value B ðn; v; AÞ. Theorem 4. Suppose F is a set of antimatroids. Then F contains only poset antimatroids if and only if the restricted Banzhaf value is the unique solution on F satisfying one-player efficiency, additivity, the necessary player property, structural monotonicity, the strong inessential player property and proxy neutrality. Proof. In a similar way as the proof of Theorem 3 it can be shown that on a set of poset antimatroids F the restricted Banzhaf value is the unique solution satisfying one-player efficiency, additivity, the necessary player property, structural monotonicity, the strong inessential player property and proxy neutrality. Suppose that A2Fis not a poset antimatroid. Define the solution g for cooperative games on antimatroids by g i ðn; u T ; AÞ ¼ 2 jp T j if i 2 P T ¼ S i2t P i; otherwise, and for arbitrary game v g i ðn; v; AÞ ¼ X d v ðt Þg i ðn; u T ; AÞ ¼ X T N ft N:i2P T g d v ðt Þ 2 : jp T j

16 62 E. Algaba et al. This solution satisfies one-player efficiency, additivity, the necessary player property, structural monotonicity, the strong inessential player property and proxy neutrality. To prove that g 6¼ B note that, if A is not a poset antimatroid then there exists j 2 N with P j 6¼ P j. By Proposition it then follows that gn; ð u T ; AÞ 6¼ B ðn; u T ; AÞ if j =2 T and ðp j np j Þ\T 6¼ ;. h An acyclic permission structure is a permission forest structure if every player has at most one predecessor. A permission forest structure is a permission tree structure if there is exactly one player i which does not have any predecessor. Algaba et al. (2, Lemma 2) showed that the permission forest structures are exactly those acyclic permission structures for which the sets of conjunctive and disjunctive feasible coalitions coincide. We also showed that the poset antimatroids satisfying the path property are exactly those antimatroids that can be obtained as the set of conjunctive or disjunctive feasible coalitions of some permission forest structure. From Theorem 4 we directly obtain a characterization of the Banzhaf value restricted to the class of poset antimatroids satisfying the path property, i.e., antimatroids that are obtained as the feasible coalitions for permission forest or tree structures. This result is interesting from an economic point of view since in economic theory we often encounter hierarchical structures that can be represented by forests or trees (see, e.g., a hierarchically structured firm). Corollary. Let F be a set of a poset antimatroids on N satisfying the path property. Then the restricted Banzhaf value is the unique solution on F satisfying one-player efficiency, additivity, the necessary player property, structural monotonicity, the strong inessential player property and proxy neutrality. 5 An application: auction situations We have already pointed out that an antimatroid can be the conjunctive feasible coalition set as well as the disjunctive feasible coalition set of some permission structure if and only if it is a poset antimatroid satisfying the path property (see Algaba et al. (2)). A special class of such antimatroids are the feasible sets of peer group situations as considered in Branzei et al. (22). In fact, they consider games with permission tree structure ðn; v; SÞ. The games v assign zero dividends to all coalitions that are not paths. The restricted peer group game then coincides with the (conjunctive or disjunctive) restricted game arising from this game with permission (tree) structure. Given that all coalitions that are not paths get a zero dividend, the restricted game is equal to the game itself. (Note that the same restricted game is obtained if we consider the game that assigns to every player i the dividend of its path P i ). Defining one-player efficiency, additivity, the necessary player property, structural monotonicity, the strong inessential player property and proxy neutrality restricted to this class we characterize the restricted Banzhaf value on this class. As argued by Branzei et al. (22) peer group situations generalize some other situations such as sealed bid second price auction situations (see Rasmusen, 994). Consider a seller of an object who has a reservation value r, and a set N ¼f;...; ng of n bidders. Each bidder has a valuation

17 An axiomatization of the Banzhaf value for cooperative games on antimatroids 63 w i r for the object. Assume that the bidders are labelled such that w > > w n. Using dominant bidding strategies for such auction situations Branzei, Fragnelli and Tijs (22) define the corresponding peer group situation that can be represented as the game with permission (tree) structure ðn; v; SÞ with SðiÞ ¼fiþg for i n, SðnÞ ¼;, and the game v determined by the dividends d v ðf;...; igþ ¼ w i w iþ if i n and d v ðnþ ¼ w n r: All other coalitions have a zero dividend. Clearly these games are determined by the set N of players and the vector of valuations ðw; rþ 2R nþ þ. Given such a valuation vector let BðN; w; rþ denote the restricted Banzhaf value of the corresponding game on the antimatroid A¼ f;; fg; f; 2g;...; f; 2;...; n g; Ng. Allowing the strict inequalities to be weak inequalities w w n, applying the axioms stated above to these situations yields that one-player efficiency straightforwardly says that f i ðfig; w; rþ ¼w i r. Structural monotonicity states that f i ðn; w; rþ f j ðn; w; rþ if w i w j. The strong inessential player property states that if w i ¼ r; then f i ðn; w; rþ ¼ and f j ðn; w; rþ ¼f j ðnnfig; w i ; rþ for all j 6¼ i; where ðw i ; rþ 2R n þ is given by ðw iþ j ¼ w j for all j 6¼ i. Specifying additivity we must take care that the underlying permission structure does not change. So, we require additivity only for reservation value vectors that preserve the order of players, i.e., f ðn; w þ z; r þ sþ ¼f ðn; w; rþþfðn; z; sþ if w i w j, z i z j. We refer to this as additivity over order preserving valuations. The neutrality axiom that we use for auction situations establishes that given h 2 NnN, f h ðn fhg; w ; rþþf j ðn fhg; w ; rþ ¼f j ðn; w; rþ if ðw ; rþ 2R nþ2 þ is given by w j ¼ w j ; w h ¼ w j and w i ¼ w i for all i 6¼ j; h. Note that this neutrality means that player j participates as representative of a hierarchical society formed by two players h; j where player h is superior to j. In this case, the value is neutral with respect to the negotiation with the superior h or with the society formed by h and j. Notice that although the auction follows the same above process where j, in some sense, is the representative of the society, the feasible coalitions do not form a chain in this case. Therefore the neutrality axiom described above is a special case of strong proxy neutrality, but not of proxy neutrality. For example, consider an auction situation with five players. In this case, the feasible coalitions are represented in Figure. Suppose that player 3 is the representative in the hierarchical society formed by 3 and 6 where 6 is superior to 3, in others words 6 controls 3. This information, leads to the rest of bidders think about the feasible coalitions of Figure 2, where besides negotiation with player 3 and therefore with the superior 6, also negotiations directly with 6 appear. Finally, we can give a characterization of the restricted Banzhaf value for auction situations without using the necessary player property. Theorem 5. The restricted Banzhaf value BðN; w; rþ is the unique solution for auction situations ðw; rþ 2R nþ þ satisfying one-player efficiency, structural monotonicity, the strong inessential player property, additivity over order preserving valuations and strong proxy neutrality. Moreover, B i ðn; w; rþ ¼ w Xn i 2i h¼iþ w h r 2h 2 n :

18 64 E. Algaba et al. {,2,3,4,5} {,2,6,3,4,5} {,2,3,4} {,2,3} {,2} {} Figure. {,2,6,3,4} {,2,6,3} {,2} Figure. 2 {,2,6,4,5} {,2,6,4} {,2,6} {} Proof. For auction situations, BðN; w; rþ is characterized by the axioms from Corollary. However, we have to prove that we do not have to use the necessary player property and need additivity only over order preserving valuations. Suppose that f is a solution for auction situations that satisfies the axioms, and let ðw; rþ 2R nþ þ be an auction situation. For k ¼ ;...; n define the auction situation ðw k ; Þ by w k i ¼ w k w kþ for all i 2f;...; kg, w k i ¼ for all i 2fkþ;...; ng, and define ðw n ; rþ by w n i ¼ w n for all i 2f;...; ng. Let k 2f;...; n g. Structural monotonicity implies that all f i ðn; w k ; Þ are equal for all i 2f;...; kg, i.e., f i ðn; w k ; Þ ¼c k ; i k for some c k 2 R. On the other hand, structural monotonicity also implies that f i ðn; w n ; rþ ¼c n, i 2 N, for some c n 2 R. The strong inessential player property implies that f i ðn; w k ; Þ ¼ for all i 2fkþ;...; ng. If N ¼fig then fðn; w; rþ is uniquely determined by one-player efficiency. Proceeding by induction assume that fðn ; w; rþ is uniquely determined if jn j < jnj. (i) For k ¼ ;...; n consider the auction situation ðw k ; Þ. Then there exist inessential players. Take an inessential player j. The strong inessential player property implies that f j N; w k ; ¼ and f i N; w k ; ¼ fi Nnfjg; w k j ; for i 2 N nfjg. Thus by the induction hypothesis f i N; w k ; is uniquely determined for all i 2 Nnfjg. (ii) For k ¼ n consider ðw n ; rþ. Then, strong proxy neutrality implies that for h; j 2 N; f h ðn; w n ; rþþf j N; w n ð ; rþ ¼ f j Nnfhg; w n h ; r : ð2þ With (2) it follows that c n ¼ 2 f j Nnfhg; w n h ; r. With the induction hypothesis c n is uniquely determined, and so is fðn; w n ; rþ. Since all ðw k ; Þ; k ¼ ;...; n, and ðw n ; rþ are order preserving, additivity over order preserving valuations determines f ðn; w; rþ. Let ðw; rþ 2R nþ þ be an auction situation. Then its corresponding poset antimatroid is A¼ f;; fg;...; f; 2;...; n g; Ng: It follows from Proposition that

19 An axiomatization of the Banzhaf value for cooperative games on antimatroids 65 B i ðw; rþ ¼ X ft N:i2P T d v ðt Þ 2 jp T j g ¼ w i w iþ 2 i þ Xn k¼iþ w k w kþ 2 k þ w n r 2 n ¼ w Xn i w k r 2i 2k 2 k¼iþ n : h From the proof of the above theorem it follows that structural monotonicity could be replaced by symmetry stating that f i ðn; w; rþ ¼f j ðn; w; rþ if w i ¼ w j. Note that this cannot be done in more general cases as discussed earlier in the paper. In a similar way we can characterize solutions for other economic situations such as airport games or hierarchically structured firms. References Algaba E, Bilbao JM, Brink R, Jiménez-Losada A (2) Cooperative games on antimatroids. CentER Discussion Paper No. 24, Tilburg University Algaba E, Bilbao JM, Brink R, Jiménez-Losada A (23) Axiomatizations of the Shapley value for cooperative games on antimatroids. Math Meth Oper Res 57:49 65 Banzhaf JF (965) Weighted Voting Doesn t Work: A Mathematical Analysis. Rutgers Law Review 9: Bilbao JM (2) Cooperative games on combinatorial structures. Kluwer Academic Publishers Borm P, Owen G, Tijs S (992) On the Position Value for Communication Situations. SIAM J Disc Math 5: Braˆ nzei R, Fragnelli V, Tijs S (22) Tree connected peer group situations and peer group games. Math Meth Oper Res 55:93 6 Brink R, Gilles RP (996) Axiomatizations of the conjunctive permission value for games with permission structures. Games and Econ Behav 2:3 26 Brink R (997) An axiomatization of the disjunctive permission value for games with a permission structure. Int J Game Theory 26:27 43 Brink R (999) An axiomatization of the conjunctive permission value for games with a hierarchical permission structure. In: Swart H (ed) Logic, Game Theory and Social Choice, pp Brink R (2) Banzhaf permission values for games with a permission structure. Technical Report 34, Department of Mathematics, University of Texas at Arlington, USA Derks J, Gilles RP (995) Hierarchical organization structures and constraints on coalition formation. Int J Game Theory 24:47 63 Derks J, Haller H (994) Null Players Out?. Working Paper, University of Limburg, Maastricht Derks J, Peters H (992) A Shapley Value for Games with Restricted Coalitions. Int J Game Theory 2: Dilworth RP (94) Lattices with unique irreducible decompositions. Ann Math 4: Dubey P, Shapley LS (979) Mathematical properties of the banzhaf power index. Math Oper Res 4:99 3 Edelman PH, Jamison RE (985) The theory of convex geometries. Geom Dedicata 9: Faigle U, Kern W (993) The Shapley value for cooperative games under precedence constraints. Int J Game Theory 2: Gilles RP, Owen G (999) Cooperative games and disjunctive permission structures. Department of Economics, Virginia Polytechnic Institute and State University, Blacksburg, Virginia Gilles RP, Owen G, Brink R (992) Games with permission structures: The Conjunctive Approach. Int J Game Theory 2: Haller H (994) Collusion properties of values. Int J Game Theory 23:26 28 Jiménez-Losada A (998) Valores para juegos sobre estructuras combinatorias. Ph.D. Dissertation at