Skip to main content
Log in

Stable coalition structures in symmetric majority games: a coincidence between myopia and farsightedness

  • Published:
Theory and Decision Aims and scope Submit manuscript

Abstract

The objective of this paper is to study stable coalition structures in symmetric majority games. We assume that players deviate from a coalition structure to another to maximize their power given by the Owen power index. We introduce three myopic core concepts and one farsighted stability concept, the farsighted vNM stable set. Our main result is that the pessimistic core, the largest myopic core, coincides with some farsighted vNM stable set for any number of players. Moreover, we show that a coalition structure belongs to the pessimistic core and the farsighted vNM stable set if and only if it contains an exact majority coalition.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Similar content being viewed by others

Notes

  1. Hart and Kurz (1983) focus on the theoretical aspect of the CS-value, whereas Hart and Kurz (1984) focus on its applications.

  2. Therefore, Kamijo (2009)’s value can be applied to the same question in the manner we elaborate from now on.

  3. For example, if \(N=\{1,2,3\}\) and \(\mathcal {P}=12|3\), then all consistent orders are 123, 213, 312, 321.

  4. To be more mathematically precise, we do not have to exclude \(S\in \mathcal {P}\) from the definition of deviation for the projective core and the pessimistic core. However, this difference does not matter in this paper.

  5. Diamantoudi and Xue (2007) consider caution as another behavioral assumption on deviating coalitions. Mauleon and Vannetelbosch (2004) also define the largest cautious consistent set, LCCS, as a refinement of the largest consistent set defined by Chwe (1994). They examine LCCS in games with positive externalities. The relationships between LCCS and our myopic cores are still open.

  6. The number of partitions such as \(\mathcal {P}'\) increases as n increases.

  7. In general, for any S with \(|S|=\frac{n}{2}\), we have \(\phi _S(\mathcal {P})=1-\frac{r!}{(r+1)!}=\frac{r}{r+1}\), where r is the number of coalitions in \(\mathcal {P}\setminus S\).

  8. This property and the following two properties are ascribed to the Owen value. For any partition \(\mathcal {P}\), any coalition \(S\in \mathcal {P}\) \((|S|\ge 2)\) and any \(i,j\in S\), we have \(\phi _i(\mathcal {P})=\phi _j(\mathcal {P})\). See Hart and Kurz (1984).

  9. For any partition \(\mathcal {P}\) and any coalition \(S,T\in \mathcal {P}\), if \(|S|=|T|\) then \(\sum _{j\in S}\phi _j(\mathcal {P})=\sum _{j\in T}\phi _j(\mathcal {P})\). See Hart and Kurz (1984).

  10. For any partition \(\mathcal {P}\) and any coalition \(S,T\in \mathcal {P}\), if \(|S|\ge |T|\) then \(\sum _{j\in S}\phi _j(\mathcal {P})\ge \sum _{j\in T}\phi _j(\mathcal {P})\). Note that this implies symmetry across coalitions. See Alonso-Meijide et al. 2009.

  11. In general, for any partition \(\mathcal {P}\) and any coalition \(S\in \mathcal {P}\) with \(|S|=k-1\), we have \(\phi _S(\mathcal {P})=\frac{|\mathcal {P}|-1}{|\mathcal {P}|}\).

  12. If \(|S|\ge k\), then the coalition S should be a winning coalition and gets 1 in \(\mathcal {P}\). Hence, \(\phi _i(\mathcal {P})\le \frac{1}{k}\) for any \(i\in S\). Since \(\phi _i(\mathcal {P}^{*})=\frac{1}{k}\) for any \(i\in K\), we have \(\mathcal {P}^{*}\succsim _i \mathcal {P}\) for any \(i\in K\cap S\).

  13. We have the following inequalities: for \(n\ge 3\),

    $$\begin{aligned} \frac{1}{k}-\frac{n-2k+3}{(n-k+1)(n-k+2)} = \left\{ \begin{array}{ll} \frac{2}{n+3}>0 &{} \text { if } n \text { is odd } (k=\frac{n+1}{2}),\\ \frac{2(n-2)}{n(n+2)} >0 &{} \text { if } n \text { is even }(k=\frac{n}{2}+1).\\ \end{array} \right. \end{aligned}$$
  14. We have the following inequalities: for \(n\ge 3\),

    $$\begin{aligned} \frac{1}{k}-\frac{1}{(n-k+1)(n-k)}= & {} \frac{1}{4k(n-k+1)(n-k)}(n+1)(n-3)\ge 0,\ (n\ge 3);\\ \frac{1}{k}-\frac{1}{2(n-k+1)(n-k)}= & {} \frac{1}{4k(n-k+1)(n-k)}(n^2-3n-2)> 0,\ (n\ge 4). \end{aligned}$$

References

  • Abe, T. (2017). Consistency and the core in games with externalities. International Journal of Game Theory. https://doi.org/10.1007/s00182-017-0581-y

    Article  Google Scholar 

  • Abe, T., & Funaki, Y. (2017). The non-emptiness of the core of a partition function form game. International Journal of Game Theory, 46, 715–736.

    Article  Google Scholar 

  • Aumann, R. J., & Drèze, J. H. (1974). Cooperative games with coalition structures. International Journal of game theory, 3, 217–237.

    Article  Google Scholar 

  • Alonso-Meijide, J. M., Bowles, C., Holler, M. J., & Napel, S. (2009). Monotonicity of power in games with a priori unions. Theory and Decision, 66, 17–37.

    Article  Google Scholar 

  • Banerjee, S., Konishi, H., & Sönmez, T. (2001). Core in a simple coalition formation game. Social Choice and Welfare, 18, 135–153.

    Article  Google Scholar 

  • Bloch, F. (1996). Sequential formation of coalitions in games with externalities and fixed payoff division. Games and Economic Behavior, 14, 90–123.

    Article  Google Scholar 

  • Bogomolnaia, A., & Jackson, M. O. (2002). The stability of hedonic coalition structures. Games and Economic Behavior, 38, 201–230.

    Article  Google Scholar 

  • Casajus, A. (2009). Outside options, component efficiency, and stability. Games and Economic Behavior, 65, 49–61.

    Article  Google Scholar 

  • Chwe, M. S. Y. (1994). Farsighted coalitional stability. Journal of Economic Theory, 63, 299–325.

    Article  Google Scholar 

  • Diamantoudi, E., & Xue, L. (2003). Farsighted stability in hedonic games. Social Choice and Welfare, 21, 39–61.

    Article  Google Scholar 

  • Diamantoudi, E., & Xue, L. (2007). Coalitions, agreements and efficiency. Journal of Economic Theory, 136, 105–125.

    Article  Google Scholar 

  • Funaki, Y., & Yamato, T. (1999). The core of an economy with a common pool resource: A partition function form approach. International Journal of Game Theory, 28, 157–171.

    Article  Google Scholar 

  • Greenberg, J. (1994). Coalition structures. Handbook of game theory with economic applications, chapter 37. Amsterdam: Elsevier.

    Google Scholar 

  • Harsanyi, J. C. (1974). An equilibrium-point interpretation of stable sets and a proposed alternative definition. Management Science, 20, 1472–1495.

    Article  Google Scholar 

  • Hart, S., & Kurz, M. (1983). Endogenous formation of coalitions. Econometrica, 51, 1047–1064.

    Article  Google Scholar 

  • Hart, S., & Kurz, M. (1984). Stable coalition structures. In M. J. Holler (Ed.), Coalitions and collective action (Vol. 51, pp. 235–258). Wüerzburg: Physica-Verlag.

    Google Scholar 

  • Kamijo, Y. (2009). A two-step Shapley value for cooperative games with coalition structures. International Game Theory Review, 11, 207–214.

    Article  Google Scholar 

  • Kóczy, L. (2007). A recursive core for partition function form games. Theory and Decision, 63, 41–51.

    Article  Google Scholar 

  • Kóczy, L., & Lauwers, L. (2004). The coalition structure core is accessible. Games and Economic Behavior, 48, 86–93.

    Article  Google Scholar 

  • Mauleon, A., & Vannetelbosch, V. (2004). Farsightedness and cautiousness in coalition formation games with positive spillovers. Theory and Decision, 56, 291–324.

    Article  Google Scholar 

  • Owen, G. (1977). Values of games with a priori unions. In Essays in mathematical economics and game theory (pp. 76–88). Berlin: Springer.

    Google Scholar 

  • Ray, D., & Vohra, R. (1997). Equilibrium binding agreements. Journal of Economic Theory, 73, 30–78.

    Article  Google Scholar 

  • Ray, D., & Vohra, R. (2015). The farsighted stable set. Econometrica, 83, 977–1011.

    Article  Google Scholar 

  • Wiese, H. (2007). Measuring the power of parties within government coalitions. International Game Theory Review, 9, 307–322.

    Article  Google Scholar 

  • Xue, L. (1997). Nonemptiness of the largest consistent set. Journal of Economic Theory, 73, 453–459.

    Article  Google Scholar 

Download references

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Takaaki Abe.

Additional information

The author is grateful to the anonymous referees and an editor for their helpful suggestions and comments. The author also thanks Yukihiko Funaki, Ryo Kawasaki, Fuhito Kojima, and Shigeo Muto for their advices. The author acknowledges financial support from the Japan Society for Promotion of Science (JSPS).

Appendix

Appendix

We first prove Proposition 2 and, next, Proposition 1.

1.1 Proof of Proposition 2

Proof

For any partition \(\mathcal {P}\), consider the following three cases:

  1. (A)

    \(\mathcal {P}\) contains a majority coalition whose size is strictly greater than k;

  2. (B)

    \(\mathcal {P}\) contains an exact majority coalition;

  3. (C)

    for every coalition in \(\mathcal {P}\), its size is strictly less than k.

We show that the partitions satisfying B are in the pessimistic core and the partitions satisfying A or C are not. Let K be an exact majority coalition. \(\square \)

Case A Let \(\mathcal {P}\) be be a partition satisfying A. Let \(K^+\in \mathcal {P}\) be a coalition such that \(|K^+|=k^+> k\). Let \(\mathcal {P}^{*}\) be be a partition satisfying B such that \(K\in \mathcal {P}^{*}\) and \(K\subseteq K^+\). Since for any \(i\in K\cap K^+=K\), \(\phi _i(\mathcal {P}^{*})=\frac{1}{k}>\frac{1}{k^+}=\phi _i(\mathcal {P})\), the players in \(K\subsetneq K^+\) have an incentive to deviate to \(\mathcal {P}^{*}\). Whatever partition the other players, \(N\setminus K\), forms, the players in K get \(\frac{1}{k}\). Thus, \(\mathcal {P}\) is not in the pessimistic core.

Case B Let \(\mathcal {P}\) be a partition satisfying B and \(K\in \mathcal {P}\). Let \(S\subseteq N\) be a deviating coalition. We typically use \(\mathcal {P}^{*}\) with \(S\in \mathcal {P}^{*}\) to denote a partition after S’s deviation. First, any coalition \(S\subseteq N\) with \(|S|\ge k\) does not deviate from \(\mathcal {P}\), because for some player \(i\in K\cap S\), \(\phi _i(\mathcal {P})=\frac{1}{k}\ge \frac{1}{|S|}=\phi _i(\mathcal {P}^{*})\) for any partition \(\mathcal {P}^{*}\) with \(S\in \mathcal {P}^{*}\). Next, consider a coalition S with \(|S|< k\). We focus on the case of \(S\cap K\ne \emptyset \), because if \(S\cap K= \emptyset \), then \(S\subseteq N\setminus K\) and any player i in S gets zero both before and after their deviation: \(\phi _i(\mathcal {P})=0=\phi _i(\mathcal {P}^{*})\).

In the case of odd n, \(|N\setminus S| \ge k\) Hence, in the pessimistic view, \(\phi _S(\mathcal {P}^{*})=0\), because the players in \(N\setminus S\) form their coalition \(N\setminus S\), and the coalition \(N\setminus S\) is a majority coalition, K or \(K^+\), in \(\mathcal {P}^{*}\). In other words, any S with \(|S|<k\) has no incentive to deviate from \(\mathcal {P}\).

In the case of even n If \(|S|\le k-2\), then it is the same as the case of odd n. If \(|S|= k-1=\frac{n}{2}\), then the power of coalition S increases as the number of coalitions except S increases.Footnote 7 Therefore, in the pessimistic view, the players in coalition S expect that the players in \(N\setminus S\) form their coalition \(\{N\setminus S\}\), where \(|S|=|N\setminus S| = \frac{n}{2}=k-1\) and \(\mathcal {P}^{*}=\{S,N\setminus S\}\). Hence, each player in coalition S expects their minimum \(\frac{1}{2}\cdot \frac{1}{|S|}\) per capita after their deviation. We have, for any \(i\in K\cap S\),

$$\begin{aligned} \phi _i(\mathcal {P})-\phi _i(\mathcal {P}^{*})= \frac{1}{k}-\frac{1}{2}\frac{1}{|S|} = \frac{2}{n+2} - \frac{1}{n} = \frac{n - 2}{n(n+2)} \ge 0, \end{aligned}$$

where the last inequality holds as \(n\ge 3\).

Case C Let \(\mathcal {P}\) be a partition satisfying C. We consider the players who obtain at least \(\frac{1}{k}\) in \(\mathcal {P}\). Let \(\hat{K}=\{j\in N | \phi _j(\mathcal {P})\ge \frac{1}{k}\}\) and \(\hat{k}=|\hat{K}|\). We must have \(\hat{k} \le k\) because of \(\sum _{j\in N}\phi _j(\mathcal {P})=1\). Moreover, we claim that \(\hat{k}\ne k\) as follows.

Claim 1

\(\hat{k}\le k-1\).

Proof

Assume that \(\hat{k}= k\). Let \(\mathcal {Z}\) be a partition of \(\hat{K}\), namely, \(\mathcal {Z}\) is a subpartition of \(\mathcal {P}\). Now, \(\mathcal {Z}\) must consist of at least two coalitions, because \(\mathcal {Z}\) is a partition of exactly k players and for every coalition in \(\mathcal {P}\) its size is strictly less than k. We consider the following order of coalitions in \(\mathcal {P}\). First, we arrange the coalitions in \(\mathcal {Z}\) in order of their size (the biggest coalition in \(\mathcal {Z}\) is located at the top) and remove (one of) the smallest coalition(s) in \(\mathcal {Z}\) from the order. At least \(\frac{k}{2}\) players are now lined up. Next, we arrange the coalitions which are not in \(\mathcal {Z}\). The number of the players who are not in \(\mathcal {Z}\) is \(n-\hat{k}=n-k=\frac{n}{2}-1\). Hence, there exists a coalition which becomes a pivot and is not in \(\mathcal {Z}\). In other words, the power index, \(\phi \), assigns a positive value to this coalition. This, however, contradicts \(\hat{k}= k\), because \(\hat{k}= k\) implies that any player not in \(\hat{K}\) gets zero.

We assume \(\hat{k}\le k-1\) hereafter.

In the case of odd n The fact \(\hat{k}\le k-1\) implies that \(n-\hat{k}\ge n-(k-1)\): in \(\mathcal {P}\), the number of players who get strictly less than \(\frac{1}{k}\) is greater than \(n-(k-1)= n-(\frac{n+1}{2}-1)=k\). Hence, these players have an incentive to form coalition S such that \(|S|=k\) and deviate from \(\mathcal {P}\).

In the case of even n If \(\hat{k}\le k-2\), then it is the same as the case of odd n. If \(\hat{k}= k-1=\frac{n}{2}\), the other \(\frac{n}{2}\) players (namely, \(N\setminus \hat{K}\)) obtain strictly less than \(\frac{1}{k}\) in \(\mathcal {P}\). We show that, in the pessimistic view, the players in \(N\setminus \hat{K}\) have an incentive to form a deviating coalition S such that \(S= N\setminus \hat{K}\) with \(|S|= n-\hat{k} = \frac{n}{2}\). If \(S=N\setminus \hat{K}\) and \(|S|=\frac{n}{2}\), then the deviating players in S expect to get \(\frac{1}{2}\frac{1}{|S|}=\frac{1}{n}\) per capita after their deviation (see the case of even n in Case B). Therefore, it suffices to show that \(\phi _i(\mathcal {P})<\frac{1}{n}\) for any \(i\in N\setminus \hat{K}\).

Claim 2

If \(\hat{k}=k-1=\frac{n}{2}\), then \(\phi _i(\mathcal {P})<\frac{1}{n}\) for any \(i\in N\setminus \hat{K}\).

Proof

Let \(T=\{i\in N\setminus \hat{K}| \phi _i(\mathcal {P})\ge \frac{1}{n}\}\) and \(t=|T|\). We have

$$\begin{aligned}&\hat{k}\cdot \frac{1}{k} + t\cdot \frac{1}{n} +0 \le \sum _{j\in \hat{K}}\phi _j(\mathcal {P}) +\sum _{j\in T}\phi _j(\mathcal {P}) + \sum _{j\in N\setminus (\hat{K}\cup T)}\phi _j(\mathcal {P})=1,\quad \text {and}\\&\hat{k}\cdot \frac{1}{k} + t\cdot \frac{1}{n} = \frac{n}{2}\cdot \frac{2}{n+2}+\frac{t}{n} = \frac{n}{n+2}+\frac{t}{n}. \end{aligned}$$

We obtain \(t\le \frac{2n}{n+2}\). As \(n\ge 3\), \(1\le \frac{2n}{n+2}< 2\). Hence, t should be 0 or 1. We assume \(t=1\) and call the player t. The player t forms his singleton coalition in \(\mathcal {P}\), because if he is in a coalition consisting of more than two players, symmetry within a coalition implies that his partners also obtain at least \(\frac{1}{n}\), which contradicts \(t=1\).Footnote 8 Moreover, similarly, symmetry across coalitions implies that the player t’s singleton coalition is the only singleton coalition in \(\mathcal {P}\).Footnote 9 In other words, the other coalitions consist of at least two players. We take a coalition \(S'\in \mathcal {P}\) such that \(|S'|\ge 2\), \(S'\subseteq N\setminus \hat{K}\). Each player in \(S'\) gets strictly less than \(\frac{1}{k}\) in \(\mathcal {P}\) because they are not the members of \(\hat{K}\). By local monotonicity and inequality \(|S'|>|\{t\}|=1\), we obtain \(\phi _{S'}(\mathcal {P})\ge \phi _t(\mathcal {P})=\frac{1}{n}\).Footnote 10 Hence, we have

$$\begin{aligned}&\hat{k}\cdot \frac{1}{k} + \frac{1}{n} + \frac{1}{n}+0 \le \sum _{j\in \hat{K}}\phi _j(\mathcal {P}) + \phi _t(\mathcal {P}) + \phi _{S'}(\mathcal {P}) +\sum _{j\not \in (\hat{K}\cup \{t\}\cup S')}\phi _j(\mathcal {P})=1,\quad \text {and}\\&\hat{k}\cdot \frac{1}{k} + \frac{1}{n} + \frac{1}{n} = \frac{n}{n+2} + \frac{2}{n} = \frac{n^2+2n+4}{n^2+2n} > 1. \end{aligned}$$

This is a contradiction. Hence, we have \(t=0\), i.e., every player in \(N\setminus \hat{K}\) gets strictly less than \(\frac{1}{n}\) in \(\mathcal {P}\). \(\square \)

Thus, in \(\mathcal {P}\), the players in \(N\setminus \hat{K}\) have an incentive to from their coalition \(N\setminus \hat{K}\) and deviate to get \(\frac{1}{n}\) even in the pessimistic view. \(\square \)

1.2 Proof of Proposition 1

Proof

For any partition \(\mathcal {P}\), consider the following three cases:

  1. (A)

    \(\mathcal {P}\) contains a majority coalition whose size is strictly greater than k;

  2. (B)

    \(\mathcal {P}\) contains an exact majority coalition;

  3. (C)

    for every coalition in \(\mathcal {P}\), its size is strictly less than k.

Let V be a set of partitions satisfying B (the set of partitions satisfying B is the pessimistic core. See Proposition 2). \(\square \)

Internal stability Let \(\mathcal {P},\mathcal {P}'\) be any two different partitions in V. Let \(K\in \mathcal {P}\) and \(K'\in \mathcal {P}'\) (\(K\ne K'\) and \(k=k'\)) be exact majority coalitions. Since K and \(K'\) are exact majority coalitions, we have \(K\cap K'\ne \emptyset \). Now, assume that there exists a path from \(\mathcal {P}\) to \(\mathcal {P}'\) satisfying the conditions of Definition 4.

As K is an exact majority coalition in \(\mathcal {P}\), we have \(\phi _i(\mathcal {P})=0\) for any \(i\in N\setminus K\) and \(\phi _i(\mathcal {P})=\frac{1}{k}\) for any \(i\in K\). Hence, in the path, the first deviating coalition \(S^1\) should be a subset of \(N\setminus K\). Let the resulting partition be \(\mathcal {P}^1\). As \(S^1\subseteq N\setminus K\), K is still in \(\mathcal {P}^1\). We have \(\phi _i(\mathcal {P}^1)=0\) for any \(i\in N\setminus K\) and \(\phi _i(\mathcal {P}^1)=\frac{1}{k}\) for any \(i\in K\). Hence, the second deviating coalition \(S^2\) is a subset of \(N\setminus K\). In general, for any step j, we have \(\phi _i(\mathcal {P}^{j})=\frac{1}{k}\) for any \(i\in K\). On the other hand, for the path to reach partition \(\mathcal {P}'\), there must be at least one coalition \(S^{j^{*}}\) and partition \(\mathcal {P}^{j^{*}}\) (at step \(j^{*}\)) such that \(K\cap K'\subseteq S^{j^{*}}\) and \(\mathcal {P}'\succ _{S^{j^{*}}}\mathcal {P}^{j^{*}}\); otherwise, \(K=K'\). However, we have \(\phi _i(\mathcal {P}^{j^{*}})=\frac{1}{k}=\frac{1}{k'}=\phi _i(\mathcal {P}')\) for any \(i\in K\cap K'\). This is a contradiction.

For any K and any two different partitions \(\mathcal {P},\mathcal {P}'\) such that \(K\in \mathcal {P}\) and \(K\in \mathcal {P}'\), we have \(\phi _i(\mathcal {P})=\phi _i(\mathcal {P}')\) for any player \(i\in N\). Thus, there is no indirect dominance within V.

External stability We show that for any partition \(\mathcal {P}\) satisfying A or C, there exists a partition \(\mathcal {P}^{*}\in V\) such that \(\mathcal {P}^{*}\) indirectly dominates \(\mathcal {P}\).

Case A For any \(\mathcal {P}\) satisfying A, let \(K^+\in \mathcal {P}\) denote a majority coalition whose size \(k^+\) is strictly greater than k. There exists a partition \(\mathcal {P}^{*}\in V\) and an exact majority coalition \(K\in \mathcal {P}\) such that \(K\subsetneq K^+\) and \(\phi _i(\mathcal {P}^{*})=\frac{1}{k}>\frac{1}{k^+}= \phi _i(\mathcal {P})\) for any \(i\in K\). Thus, \(\mathcal {P}^{*}\) (in)directly dominates \(\mathcal {P}\).

Case C Let \(\mathcal {P}\) be a partition satisfying C. We consider the set of players who obtain at least \(\frac{1}{k}\) in \(\mathcal {P}\). Let \(\hat{K}=\{j\in N | \phi _j(\mathcal {P})\ge \frac{1}{k}\}\) and \(\hat{k}=|\hat{K}|\). As Claim 1 in the proof of Proposition 2, we have \(\hat{k} \le k-1\).

In the case of odd n \(\hat{k} \le k-1\) means that at least k players get strictly less than \(\frac{1}{k}\) in \(\mathcal {P}\). They form coalition S consisting of k players and directly deviate to partition \(\{S\}\cup (\mathcal {P}|_{N\setminus S})\in V\).

In the case of even n If \(\hat{k}\le k-2\), then it is the same as the case of odd n. If \(\hat{k}= k-1=\frac{n}{2}\), then the other \(\frac{n}{2}\) players (namely, \(N\setminus \hat{K}\)) obtain strictly less than \(\frac{1}{k}\). Let \(\mathcal {Q}\) be a partition of \(N\setminus \hat{K}\), namely \(\mathcal {Q}\) is a subpartition of \(\mathcal {P}\). We first show the following claim.

Claim 3

\(|\mathcal {Q}|\ge 2\).

Proof

Assume that \(|\mathcal {Q}|=1\) (i.e., \(\mathcal {Q}=\{N\setminus \hat{K}\}\), which contains \(k-1\) players). We have

$$\begin{aligned} \phi _{N\setminus \hat{K}}(\mathcal {P})=1-(|\mathcal {P}|-1)\frac{(|\mathcal {P}|-2)!}{|\mathcal {P}|!}=1-(|\mathcal {P}|-1)\frac{1}{|\mathcal {P}|(|\mathcal {P}|-1)} =\frac{|\mathcal {P}|-1}{|\mathcal {P}|}, \end{aligned}$$
(5)

for \(2\le |\mathcal {P}| \le k\).Footnote 11 This attains its minimum at \(|\mathcal {P}|=2\). Hence, \(\phi _{N\setminus \hat{K}}(\mathcal {P})\ge \frac{1}{2}\). However, since \(\phi _i(\mathcal {P})\ge \frac{1}{k}\) for any \(i\in \hat{K}\), we have \(\phi _{N\setminus \hat{K}}(\mathcal {P})=1-\phi _{\hat{K}}(\mathcal {P})\le 1-(\frac{1}{k}\cdot \hat{k})=1-(\frac{k-1}{k})=\frac{1}{k}\), which contradicts \(\phi _{N\setminus \hat{K}}(\mathcal {P})\ge \frac{1}{2}\) as \(n\ge 4\) (\(k\ge 3\) when \(n\ge 4\), and 4 is the smallest even n). \(\square \)

Next, we construct a path from \(\mathcal {P}\) to \(\mathcal {P}^{*}\in V\) via a partition \(\bar{\mathcal {P}}\) given by

$$\begin{aligned} \bar{\mathcal {P}}= & {} (\mathcal {P}|_{\hat{K}}) \cup \{N\setminus \hat{K}\}, \\ \mathcal {P}^{*}= & {} (\mathcal {P}|_{\hat{K}\setminus \{i^{*}\}}) \cup \{(N\setminus \hat{K})\cup \{i^{*}\}\}, \end{aligned}$$

where \(i^{*}\) is any player in \(\hat{K}\). Note that \(\mathcal {P}^{*} \in V\) because coalition \((N\setminus \hat{K})\cup \{i^{*}\}\) is an exact majority.

The first deviating coalition (from \(\mathcal {P}\) to \(\bar{\mathcal {P}}\)) is \(N\setminus \hat{K}\). For every player \(i\in N\setminus \hat{K}\), \(\phi _i(\mathcal {P}^{*}) > \phi _i(\mathcal {P})\), because \(\phi _i(\mathcal {P})<\frac{1}{k}\) and \(\phi _i(\mathcal {P}^{*})=\frac{1}{k}\). Moreover, by \(|\mathcal {Q}|\ge 2\), all players in \(\mathcal {Q}\) can form one coalition \(\{N\setminus \hat{K}\}\) and move to \(\bar{\mathcal {P}}\).

The second deviating coalition (from \(\bar{\mathcal {P}}\) to \(\mathcal {P}^{*}\)) is \((N\setminus \hat{K})\cup \{i^{*}\}\). We show that for any \(i\in N\setminus \hat{K}\), \(\phi _i(\bar{\mathcal {P}})<\frac{1}{k}\) (for \(i^{*}\); see Claim 5).

Claim 4

In \(\bar{\mathcal {P}}\), for any \(i\in N\setminus \hat{K}\), \(\phi _i(\bar{\mathcal {P}})<\frac{1}{k}=\phi _i(\mathcal {P}^{*})\).

Proof

We first show that \(\phi _i(\bar{\mathcal {P}})\le \frac{1}{k}\). We have

$$\begin{aligned} \frac{1}{k}-\phi _i(\bar{\mathcal {P}}) =\frac{1}{k}-\frac{|\bar{\mathcal {P}}|-1}{|\bar{\mathcal {P}}|}\cdot \frac{1}{k-1} =\frac{1}{k(k-1)}\left( \frac{k}{|\bar{\mathcal {P}}|}-1 \right) \ge 0, \end{aligned}$$

where the first equality holds in the same manner with (5) and the final inequality follows from \(k\ge |\bar{\mathcal {P}}|\). Hence, if \(k= |\bar{\mathcal {P}}|\), then \(\phi _i(\bar{\mathcal {P}})=\frac{1}{k}\). Below, we show that this case does not occur. Assume that \(|\bar{\mathcal {P}}|=k\). Note that \(\hat{k}= k-1\). From \(\bar{\mathcal {P}}= (\mathcal {P}|_{\hat{K}}) \cup \{N\setminus \hat{K}\}\), it follows that \(\mathcal {P}|_{\hat{K}}\) is a partition of \(\hat{K}\) into \(\hat{k}\) singletons, which means that the players in \(\hat{K}\) are singletons in the first partition \(\mathcal {P}\), because the first deviating coalition is \(\{N\setminus \hat{K}\}\). The local monotonicity of the Owen power index implies that for any \(i\in \hat{K}\) and any \(j\in N\setminus \hat{K}\), \(\phi _i(\mathcal {P}) \le \phi _j(\mathcal {P})\), because \(i\in \hat{K}\) forms a singleton coalition in \(\mathcal {P}\) and \(j\in N\setminus \hat{K}\) belongs to a coalition consisting of at least one player in \(\mathcal {P}\). This contradicts \(\hat{K}\), namely, \(\phi _i(\mathcal {P})\ge \frac{1}{k}\) for \(i\in \hat{K}\) and \(\phi _j(\mathcal {P})< \frac{1}{k}\) for \(j\in N\setminus \hat{K}\).

We next show that in \(\bar{\mathcal {P}}\), there exists \(i^{*}\in \hat{K}\) which agrees with the second deviation.

Claim 5

In \(\bar{\mathcal {P}}\), there exists \(i^{*}\in \hat{K}\) such that \(\phi _{i^{*}}(\bar{\mathcal {P}}) < \frac{1}{k}=\phi _{i^{*}}(\mathcal {P}^{*})\).

Proof

Assume that for every \(i\in \hat{K}\), \(\phi _{i}(\bar{\mathcal {P}}) \ge \frac{1}{k}\). We have \(\phi _{N\setminus \hat{K}}(\bar{\mathcal {P}})=\frac{|\bar{\mathcal {P}}|-1}{|\bar{\mathcal {P}}|}\) (\(2\le |\bar{\mathcal {P}}| \le k\)) in the same manner as (5). This attains its minimum \(\frac{1}{2}\) at \(|\bar{\mathcal {P}}|=2\) (namely, \(\phi _{N\setminus \hat{K}}(\bar{\mathcal {P}})=\sum _{j\in N\setminus \hat{K}}\phi _j(\bar{\mathcal {P}})\ge \frac{1}{2}\)). Hence,

$$\begin{aligned} \sum _{j\in N}\phi _j(\bar{\mathcal {P}})= \sum _{j\in \hat{K}}\phi _j(\bar{\mathcal {P}})+\sum _{j\in N\setminus \hat{K}}\phi _j(\bar{\mathcal {P}}) \ge (k-1)\frac{1}{k}+\frac{1}{2} =\frac{3}{2}-\frac{1}{k}. \end{aligned}$$

This, however, contradicts \(\sum _{j\in N}\phi _j(\bar{\mathcal {P}})=1\) as \(k\ge 3\) (i.e., \(n\ge 4\)). \(\square \)

Thus, for every player, \(i\in (N\setminus \hat{K})\cup \{i^{*}\}\), \(\phi _i(\mathcal {P}^{*}) > \phi _i(\bar{\mathcal {P}})\). \(\square \)

1.3 Proof of Proposition 3

Proof

We show that \(\mathcal {P}^{*}=\{K\}\cup [N\setminus K]\) is in the projective core. We consider the following two cases: odd n and even n. Let S be a deviating coalition and \(\mathcal {P}\) be the partition after S’s deviation; formally,

$$\begin{aligned} \mathcal {P}:=\{K\setminus S\} \cup \{S\} \cup [(N\setminus K)\setminus S]. \end{aligned}$$
(6)

Odd, condition (i) We show that for any \(S\subseteq N\) such that \(S\cap K\ne \emptyset \) and \(|S|\le k-1\), we have \(\phi _i(\mathcal {P}^{*}) \ge \phi _i(\mathcal {P})\).

For any player \(i\in S\cap K\).Footnote 12 To compute \(\phi _S(\mathcal {P})\), let \(l=|S\cap K|\) and \(r=|(N\setminus K)\setminus S|\). Note that

$$\begin{aligned} 1\le & {} |S|\le k-1, \end{aligned}$$
(7)
$$\begin{aligned} 1\le & {} l \le k-1,\nonumber \\ 0\le & {} r \le n-k. \end{aligned}$$
(8)

The number of orders that S becomes a pivot as a coalition in \(\mathcal {P}\) (formally, a pivot in its corresponding weighted majority game) is

$$\begin{aligned} \sum _{j=0}^{l-1} \left[ (1+j)! (r-j)! \left( \begin{array}{cc} r \\ j \\ \end{array}\right) \times 2 \right]= & {} 2\sum _{j=0}^{l-1} [ (j+1) r! ]\nonumber \\= & {} 2r!\sum _{j=1}^{l}j\nonumber \\= & {} r! l(1+l). \end{aligned}$$
(9)

The number of all orders (of coalitions in \(\mathcal {P}\)) is \((r+2)!\). Hence, by (9), we have

$$\begin{aligned} \phi _S(\mathcal {P})=\frac{r! l(1+l)}{(r+2)!} = \frac{l(1+l)}{(r+2)(r+1)}. \end{aligned}$$

Since \(|S|=l+n-k-r\), we have, for any \(i\in S\),

$$\begin{aligned} \phi _i(\mathcal {P}) =\frac{l(1+l)}{(r+2)(r+1)}\cdot \frac{1}{l+n-k-r}. \end{aligned}$$
(10)

This increases with respect to l. By \(|S|=l+n-k-r\), the right hand side of (7) implies \(l\le 2k+r-1-n\). By \(k=\frac{n+1}{2}\), we have \(l\le r\). Hence, with respect to l, (10) attains its maximum at \(l=r\). Assuming \(l=r\), (10) is

$$\begin{aligned} \frac{r(1+r)}{(r+2)(r+1)}\cdot \frac{1}{r+n-k-r}=\frac{r}{r+2}\cdot \frac{1}{n-k}. \end{aligned}$$
(11)

This increases with respect to r. By (8), substituting \(r=n-k\) for (11), we have

$$\begin{aligned} \frac{1}{n-k+2}= & {} \frac{2}{n+3} < \frac{2}{n+1} = \frac{1}{k}. \end{aligned}$$
(12)

Thus, (10)–(12) imply that for any \(i\in S\cap K\), \(\phi _i(\mathcal {P})<\frac{1}{k}=\phi _i(\mathcal {P}^{*})\).

Odd, condition (ii) For any \(S\subseteq N\setminus K\), by (6), we have \(\mathcal {P}=\{K\}\cup \{S\}\cup [(N\setminus K)\setminus S]\). Since K is an exact majority coalition, we have \(\phi _i(\mathcal {P}^{*})=0=\phi _i(\mathcal {P})\) for any i with \(i\in S\subseteq N\setminus K\).

Even, condition (i) The approach is the same as the case of odd n. We have (7)–(8). We change the computation of (9) as follows:

$$\begin{aligned}&\sum _{j=0}^{l-2} \left[ (1+j)! (r-j)! \left( \begin{array}{cc} r \\ j \\ \end{array}\right) \times 2 \right] + l!(r-l+1)!\left( \begin{array}{cc} r \\ l-1\\ \end{array}\right) \nonumber \\&\quad = 2\sum _{j=0}^{l-2}\left[ (j+1) r! \right] + l\cdot r!\nonumber \\&\quad = r! \cdot l^2. \end{aligned}$$
(13)

The number of all orders is \((r+2)!\). Hence, we obtain

$$\begin{aligned} \phi _S(\mathcal {P})=\frac{r! \cdot l^2.}{(r+2)!} = \frac{l^2}{(r+2)(r+1)}. \end{aligned}$$

Since \(|S|=(l+n-k-r)\), for any \(i\in S\),

$$\begin{aligned} \phi _i(\mathcal {P}) =\frac{l^2}{(r+2)(r+1)}\cdot \frac{1}{l+n-k-r}. \end{aligned}$$
(14)

This increases with respect to l. From \(|S|=l+n-k-r\) and (7), we have \(l\le 2k+r-1-n\). Moreover, because of \(k=\frac{n}{2}+1\), we obtain \(l\le r+1\). Hence, with respect to l, (14) attains its maximum at \(l=r+1\). Assuming \(l=r+1\), (14) is

$$\begin{aligned} \frac{(r+1)^2}{(r+2)(r+1)}\cdot \frac{1}{r+1+n-k-r}=\frac{r+1}{r+2}\cdot \frac{1}{n-k+1}. \end{aligned}$$
(15)

This increases with respect to r. By (8), substituting \(r=n-k\) for (15), we have

$$\begin{aligned} \frac{1}{n-k+2}= & {} \frac{2}{n+2} < \frac{2}{n+1} = \frac{1}{k}. \end{aligned}$$
(16)

Thus, from (14)–(16), we have the same conclusion as the case of odd n.

Even, condition (ii) This part is exactly the same as the case of odd n. \(\square \)

1.4 Proof of Proposition 6

Proof

We first consider the partition \(\{K\} \cup [N\setminus K]\). Although every player in \([N\setminus K]\) has an incentive to join coalition K, all players in K refuse it. Moreover, every player in \([N\setminus K]\) has no incentive to merge with any other player in \([N\setminus K]\), because he gets zero even after the merge. Every player in K has no incentive to deviate alone from K, because he gets \(\frac{n-2k+3}{(n-k+1)(n-k+2)}\ (<\frac{1}{k})\) after his deviation.Footnote 13 Every player in K has no incentive to merge with any player in \([N\setminus K]\), because he gets \(\frac{1}{(n-k+1)(n-k)}\) (\(\le \frac{1}{k}\) for \(n\ge 3\), odd) or \(\frac{1}{2(n-k+1)(n-k)}\) (\(<\frac{1}{k}\) for \(n\ge 4\), even) after the merge.Footnote 14 Next, consider a partition which contains a winning coalition whose size is strictly greater than k. The partition \(\{N\}\) is individually stable because it is Nash stable. For a partition which contains a winning coalition \(K^+\) with \(k^+>k\), every player i in \(K^+\) has no incentive to deviate from \(K^+\) because \(K^+\setminus \{i\}\) is still winning. Although every singleton player in \(N\setminus K^+\) has an incentive to join coalition \(K^+\), all players in \(K^+\) refuse it. \(\square \)

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Abe, T. Stable coalition structures in symmetric majority games: a coincidence between myopia and farsightedness. Theory Decis 85, 353–374 (2018). https://doi.org/10.1007/s11238-018-9655-y

Download citation

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s11238-018-9655-y

Keywords

Navigation