Skip to main content
Log in

Triple-consistent social choice and the majority rule

  • Original Paper
  • Published:
TOP Aims and scope Submit manuscript

Abstract

We define generalized (preference) domains \(\mathcal{D}\) as subsets of the hypercube {−1,1}D, where each of the D coordinates relates to a yes-no issue. Given a finite set of n individuals, a profile assigns each individual to an element of \(\mathcal{D}\). We prove that, for any domain \(\mathcal{D}\), the outcome of issue-wise majority voting φ m belongs to \(\mathcal{D}\) at any profile where φ m is well-defined if and only if this is true when φ m is applied to any profile involving only 3 elements of \(\mathcal{D}\). We call this property triple-consistency. We characterize the class of anonymous issue-wise voting rules that are triple-consistent, and give several interpretations of the result, each being related to a specific collective choice problem.

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. A Condorcet domain is maximal if it is no longer Condorcet when completed by any additional order.

  2. An Arrovian aggregation rule is a mapping φ from a set of profiles π of linear orders over a set \(\mathcal{A}\) of alternatives to the set of all linear orders over \(\mathcal{A}\) which satisfies:

    • If all preferences in π rank \(a\in \mathcal{A}\) above \(a^{\prime }\in \mathcal{A}\), the so does φ(π) (unanimity).

    • If two profiles π and π′ coincide in restriction to a pair {a,a′} of alternatives, then φ(π) and φ(π′) rank a and a′ in the same way (independence of irrelevant alternatives).

  3. In particular, it relates to the problem of judgment aggregation, which has generated a vast literature (see List and Puppe 2009 for a survey). We provide below other interpretations of generalized domains.

  4. In the Arrovian case of preference aggregation, where issues d are pairs of alternatives, independence implies the property of independence of irrelevant alternatives. Independence for \((\mathcal{D},n)\) is equivalently defined as follows: ∃f:⋃ n≥3{−1,1}n→{−1,1} such that for all \(\pi \in \mathcal{D}_{\varphi }^{n}\), φ(π)=(f(π1),…,f(π D )).

  5. An anonymous and independent aggregation rule φ is monotonic if \(\forall \mathcal{D}\in \varDelta \), ∀n≥3, \(\forall \pi \in \mathcal{D}_{\varphi }^{n}\), ∀d∈{1,…,D}, ∀a∈{−1,1}, [φ(π)]d=a only if \([\varphi (\widetilde{\pi })]^{d}=a\) for any \(\widetilde{\pi }\in \mathcal{D}_{\varphi }^{n}\) such that \(n_{a}(\widetilde{\pi }\mid _{d})>n_{a}(\pi \mid _{d})\).

  6. A possible interpretation of this example is that a program is a list of potential public expenses in a society facing a budget constraint that allows to undertake exactly two expenses. Individuals claim for different expenses under the commonly known budget constraint. By using any of the three rules φ m , \(\varphi _{m}^{-}\) and φ u , the society may fail at meeting the constraint.

  7. Note that we do not retain this restriction in our definition of a domain.

  8. Note that if φ is independent, then φ is almost constant if there exists a∈{−1,1} such that for all \(\mathcal{D}\in \varDelta \), for all n≥3, and for all \(\pi \in \mathcal{D}^{n}\), we have 0<n a (π)<n⇒[φ(π)]d=a.

  9. In particular, every anonymous issue-wise voting rule may select a Pareto-dominated outcome as long as any separable preference on possible outcomes is admissible (Özkal-Sanver and Sanver 2006; Benoît and Kornhauser 2010; Cuhadaroglu and Lainé 2012). Furthermore, under additively separable preferences over outcomes, the outcome of issue-wise majority voting may be less preferred by a majority of voters than another one (see Bezembinder and Van Acker 1985; Deb and Kelsey 1987, and Laffond and Lainé 2006, 2009), or may disagree with the majority will on a majority of issues (Anscombe 1976; Laffond and Lainé 2013).

  10. However, type populations in the stable club may change over time since admission juries are formed from the set of current members. Hence, stability does not mean constant distribution of types. See Barberà et al. (2001), and Granot et al. (2002) on a related problematic.

References

  • Anscombe GEM (1976) On the frustration of the majority by fulfillment of the majority’s will. Analysis 36:161–168

    Article  Google Scholar 

  • Barberà S, Maschler M, Shalev J (2001) Voting for voters: A model of electoral evolution. Games Econ Behav 37:40–78

    Article  Google Scholar 

  • Benoît JP, Kornhauser LA (2010) Only dictatorship is efficient or neutral. Games Econ Behav 70:261–270

    Article  Google Scholar 

  • Bezembinder T, Van Acker P (1985) The Ostrogorski paradox and its relation to non-transitive choice. J Math Psychol 11:131–158

    Google Scholar 

  • Black D (1948) On the rationale of group decision-making. J Polit Econ 56:23–34

    Article  Google Scholar 

  • Brams S, Kilgour D, Zwicker W (1998) The paradox of multiple elections. Soc Choice Welf 15:211–236

    Article  Google Scholar 

  • Cuhadaroglu T, Lainé J (2012) Pareto efficiency in multiple referendum. Theory Decis 72:525–536

    Article  Google Scholar 

  • Deb R, Kelsey D (1987) On constructing a generalized Ostrogorski paradox: Necessary and sufficient conditions. Math Soc Sci 14:161–174

    Article  Google Scholar 

  • Dietrich F (2006) Judgment aggregation: (im)possibility theorems. J Econ Theory 126:286–298

    Article  Google Scholar 

  • Dokow E, Holzman R (2010) Aggregation of binary evaluations. J Econ Theory 145:495–511

    Article  Google Scholar 

  • Fishburn PC (1997) Acyclic sets of linear orders. Soc Choice Welf 14:113–124

    Article  Google Scholar 

  • Fishburn PC (2002) Acyclic sets of linear orders: A progress report. Soc Choice Welf 19:431–447

    Article  Google Scholar 

  • Gaertner W (1979) An analysis and comparison of several necessary and sufficient conditions for transitivity under the majority rule. In: Laffont JJ (ed) Aggregation and revelation of preferences. North-Holland, Amsterdam

    Google Scholar 

  • Granot D, Maschler M, Shalev J (2002) Voting for voters: The unanimity case. Int J Game Theory 31:55–202

    Google Scholar 

  • Inada KI (1969) The simple majority decision rule. Econometrica 32:490–506

    Article  Google Scholar 

  • Kornhauser LA, Sager LG (1986) Unpacking the court. Yale Law J 96:82–117

    Article  Google Scholar 

  • Kornhauser LA, Sager LG (2004) Group choice in paradoxical cases. Philos Public Aff 32:249–276

    Article  Google Scholar 

  • Laffond G, Lainé J (2006) Single-switch preferences and the Ostrogorski paradox. Math Soc Sci 52:49–66

    Article  Google Scholar 

  • Laffond G, Lainé J (2009) Condorcet choice and the Ostrogorski paradox. Soc Choice Welf 32:317–333

    Article  Google Scholar 

  • Laffond G, Lainé J (2013) Unanimity and the Anscombe’s paradox. TOP 21(3):590–611

    Article  Google Scholar 

  • List C, Pettit P (2002) Aggregating sets of judgments: An impossibility result. Econ Philos 18:89–110

    Google Scholar 

  • List C, Puppe C (2009) Judgment aggregation: A survey. In: Anand P, Pattanaik P, Puppe C (eds) Handbook of rational and social choice. Oxford University Press, Oxford

    Google Scholar 

  • Maskin E (1995) Majority rule, social welfare functions, and game forms. In: Basu K, Pattanaik P, Suzumura K (eds) Choice, welfare, and development (essays in honor of Amartya Sen). Oxford University Press, Oxford, pp 100–109

    Google Scholar 

  • Monjardet B (2010) Acyclic domains of linear orders: A survey. In: Brams S, Gehrlein WV, Roberts FS (eds) The mathematics of preference, choice and order. Essays in honor of Peter C Fishburn (Studies in Social Choice and Welfare). Springer, Berlin, pp 139–160

    Google Scholar 

  • Nehring K, Puppe C (2010) Abstract Arrovian aggregation. J Econ Theory 145:467–494

    Article  Google Scholar 

  • Özkal-Sanver I, Sanver MR (2006) Ensuring Pareto optimality by referendum voting. Soc Choice Welf 27:211–219

    Article  Google Scholar 

  • Scarsini M (1998) A strong paradox of multiple elections. Soc Choice Welf 15:237–238

    Article  Google Scholar 

  • Sen AK, Pattanaik PK (1969) Necessary and sufficient conditions for rational choice under majority decision. J Econ Theory 1:178–202

    Article  Google Scholar 

  • Wilson R (1975) On the theory of aggregation. J Econ Theory 10:89–99

    Article  Google Scholar 

Download references

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Jean Lainé.

Additional information

Authors are indebted to Remzi Sanver for helpful remarks. They are also grateful to two anonymous reviewers for highly valuable comments and suggestions.

Appendix: Proof of Theorem 3

Appendix: Proof of Theorem 3

The sufficiency part follows from Theorem 2 together with Propositions 1 and 2. We prove the necessary part through several intermediate lemmas. Let φ be an anonymous and independent aggregation rule. Independence together with anonymity implies that for any domain \(\mathcal{D}\), for any issue d∈{1,…,D}, for any n≥3 and for any \((\mathcal{D},n)\)-profile \(\pi \in \mathcal{D}_{\varphi }^{n}\), [φ(π)]d only depends on n 1(π d ). Consider the set \(\mathcal{D}_{\varphi }^{3}\) of all admissible profiles which involve exactly 3 individuals. Pick any \(\pi \in \mathcal{D}_{\varphi }^{3}\) together with an issue d. Then, there exists a d ∈{−1,1} such that for all issues d,

Case 1::

\([n_{a_{d}}(\pi \mid _{d})=2\mbox{ and }n_{-a_{d}}(\pi \mid _{d})=1\Rightarrow [ \varphi (\pi )]^{d}=a_{d}]\) (property M +);

Case 2::

\([n_{a_{d}}(\pi \mid _{d})=2\mbox{ and }n_{-a_{d}}(\pi \mid _{d})=1\Rightarrow [ \varphi (\pi )]^{d}=-a_{d}]\) (property M );

Case 3::

\([n_{-a_{d}}(\pi \mid _{d})\in \{1,2\}\Rightarrow [ \varphi (\pi )]^{d}=a_{d}]\) (property M =).

Suppose that for some domain \(\mathcal{D}\), φ violates the majority criterion at some issue d. Thus there exist h>h′>0, d∈{1,…,D}, a d ∈{−1,1} and \(\pi \in \mathcal{D}_{\varphi }^{h+h^{\prime }}\) such that \(n_{a_{d}}(\pi \mid _{d})=h\), \(n_{-a_{d}}(\pi \mid _{d})=h^{\prime }\) and [φ(π)]d=−a d . We denote by h φ the smallest integer h for which the above situation holds for φ.

Lemma 1

If φφ m is triple-consistent, then h φ =2.

Proof

Suppose that for some domain \(\widetilde{\mathcal{D}}\) there exists an admissible \((\widetilde{\mathcal{D}},h_{\varphi }+h^{\prime })\)-profile \(\widetilde{\pi }\) such that \(n_{a_{d}}(\widetilde{\pi }\mid _{d})=h_{\varphi }\geq 3\), \(n_{-a_{d}}(\widetilde{\pi }\mid _{d})=h^{\prime }\) and \([\varphi (\widetilde{\pi })]^{d}=-a_{d}\) for some issue d. Write a=a d . Define \(\mathcal{D}=\{x,y,z\}\in \{-1,1\}^{3}\), where x=(−a,a,−a), y=(a,−a,a), and z=(a,a,a). Consider the \((\mathcal{D},3)\)-profile π=(x,y,z). Since n a (π d )<h φ for all d, then φ(π)=z, and thus \(\mathcal{D}\) is 3-stable under φ. Consider the \((\mathcal{D},h_{\varphi }+h^{\prime })\)-profile \(\pi ^{\prime }=(\overset{h^{\prime }}{\overbrace{x,\ldots,x}},\overset{h^{\prime }}{\overbrace{y,\ldots,y}},\overset{h_{\varphi }-h^{\prime }}{\overbrace{z,\ldots,z}})\). Since for all issues d, \(n_{a}(\pi ^{\prime }\mid _{d})=h_{\varphi }=n_{a}(\widetilde{\pi }\mid _{d})\) and \(n_{-a}(\pi ^{\prime }\mid _{d}) =h^{\prime }=n_{-a}(\widetilde{\pi }\mid _{d})\), then \(\pi ^{\prime }\in \mathcal{D}_{\varphi }^{h_{\varphi }+h^{\prime }}\). Moreover, [φ(π′)]d=−a. It follows from independence that \(\varphi (\pi ^{\prime })=(-a,-a,-a)\notin \mathcal{D}\), and therefore φ is not triple-consistent. Thus, h φ ≤2. Finally, since h φ >h′>0, then h φ =2. □

An immediate consequence of Lemma 1 is:

Lemma 2

If φ is triple-consistent and such that property M + holds, then for all \(\mathcal{D}\in \varDelta \), for all n≥3, and for all non-unanimous profiles \(\pi \in \mathcal{D}^{n}\), φ(π)=φ m (π).

We generalize below Lemma 2 to any admissible profile, unanimous or not. An intermediate step is:

Lemma 3

If φ is triple-consistent and unanimous in restriction to 3-program profiles, then φ is unanimous.

Proof

Define \(\mathcal{D}=\{x,y,z,t\}\subset \{-1,1\}^{3}\), where x=(a,a,a), y=(−a,−a,a), z=(a,−a,a), and t=(−a,a,a). Since \(\mathcal{D}\) contains all programs with a as third coordinate, and since φ is unanimous in restriction to \(\mathcal{D}_{\varphi }^{3}\), then \(\mathcal{D}\) is 3-stable under φ. Thus, triple-consistency ensures that \(\mathcal{D} \) is stable under φ. Suppose that \(n_{a}(\widetilde{\pi }\mid _{d})=n\) and \([\varphi (\widetilde{\pi })]^{d}=-a\) for some \(\widetilde{\mathcal{D}}\in \varDelta \), some n≥3 and some \(\widetilde{\pi }\in \widetilde{\mathcal{D}}_{\varphi }^{n}\). It follows from anonymity together with independence that [φ(π)]d=−a for any \(\pi \in \mathcal{D}_{\varphi }^{n}\) such that n a (π d )=n. Thus \(\mathcal{D}\) is not stable under φ, in contradiction with triple-consistency. □

Lemma 4

If φ is triple-consistent and satisfies property M +, then φ=φ m .

Proof

From Lemma 2, together with Lemma 3, it suffices to prove that φ is unanimous at all profiles \(\pi \in \mathcal{D}_{\varphi }^{3}\). Suppose the contrary. It follows from anonymity and independence that there exists a∈{1,−1} such that for all \(\pi \in \mathcal{D}_{\varphi }^{3}\) and all d, n a (π d )=3 and [φ(π)]d=−a (Property A). Define \(\mathcal{D}=\{x,y,z,t,-x,-y, -z,-t\}\subset \{-1,1\}^{4}\), where x=(a,a,a,−a), y=(a,a,−a,a), z=(a,−a,a,a), and t=(−a,a,a,a). From M +, together with Property A, we get that φ(x,y,z)=t, φ(x,y,t)=z, φ(x,z,t)=y, and φ(y,z,t)=x. Furthermore, for any ww′≠w′′∈{x,y,z,t}, we have φ(w,w,w′)=−w′=φ(−w,−w,w′), φ(w,−w,w′)=w′, φ(w,−w,−w′)=−w′, \(\varphi (w,w^{\prime },w^{\prime \prime })=\varphi (-w,-w^{\prime },w^{\prime \prime })\in \mathcal{D}\), and \(\varphi (-w,-w^{\prime }, -w^{\prime \prime })=-\varphi (w,w^{\prime },w^{\prime \prime })\in \mathcal{D}\). Thus \(\mathcal{D}\) is 3-stable under φ. Finally, we get from Lemma 2 that \(\varphi (x,y,z,t)=(a,a,a,a)\notin \mathcal{D}\). Thus, φ is not triple-consistent, a contradiction. □

We now turn to aggregation rules sharing property M =.

Lemma 5

If φ is triple-consistent and such that property M = holds, then φ is almost constant.

Proof

Consider the domain \(\mathcal{D}^{\ast }\subset \{-1,1\}^{h+1}\) defined in Table 1, where issue-wise positions appear in columns and programs in rows.

Table 1 Domain \(\mathcal{D}^{\ast }\)

Then, we define \(\mathcal{D}=\{1,-1\}^{h+1}-\{(-x_{h+2})\}\). Suppose that there exist a∈{−1,1} and \(h\neq h^{\prime } \in \mathbb{N} \) such that h+h′>3 and, for some d and some \(\pi \in \mathcal{D}_{\varphi }^{h+h^{\prime }}\), n a (π d )=h′, n a (π d )=h, and [φ(π)]d=−a. Pick up any triple \(\{x,y,z\}\subset \mathcal{D}\) such that profile \((x,y,z) \in \mathcal{D}_{\varphi }^{3}\). From the definition of \(\mathcal{D}\), we have that for all d, there exists w∈{x,y,z} such that w d=−a. Hence, property M = implies that φ(x,y,z)≠(−x h+2). Therefore \(\varphi (x,y,z)\in \mathcal{D}\), and \(\mathcal{D}\) is 3-stable under φ. Finally, define the \((\mathcal{D},h+h^{\prime })\)-profile \(\pi ^{\ast }=(x_{1},x_{2},\ldots,x_{h+1},\overset{h^{\prime }-1}{\overbrace{x_{h+2},\ldots,x_{h+2}}})\) that contains only programs in \(\mathcal{D}^{\ast }\subset \mathcal{D}\). Since ∀d∈{1,…,h+1}, then n a (π d )=h′ and n a (π d )=h. Since n a (π d )=n a (π d )=h′ and n a (π d )=n a (π d )=h, then \(\pi ^{\ast }\in \mathcal{D}_{\varphi }^{h+h^{\prime }}\). It follows from anonymity together with independence that \(\varphi (\pi ^{\ast })=-x_{h+2}\notin \mathcal{D}\). Thus, \(\mathcal{D}\) is not stable under φ, in contradiction with triple-consistency. Therefore, we have shown that \(\forall \mathcal{D}\in \varDelta \), ∀n≥3, \(\forall \pi \in \mathcal{D}_{\varphi }^{n}\), ∀a∈{−1,1}, 0<n a (π d )<n⇒[φ(π)]d=a, and thus φ is almost constant. □

It remains to consider property M .

Lemma 6

If φ is triple-consistent and such that property M holds, then \(\varphi (\pi )=\varphi _{m}^{-}(\pi )\) for all \(\mathcal{D}\in \varDelta \) and all \(\pi \in \mathcal{D}_{\varphi }^{3}\) .

Proof

It suffices to prove that ∀a∈{1,−1}, \(\forall \pi \in \mathcal{D}_{\varphi }^{3}\), ∀d, n a (π d )=3⇒[φ(π)]d=−a. Suppose the contrary, and consider again the domain \(\mathcal{D}\) defined in the proof of Lemma 4, where \(\forall w\in \mathcal{D}\), \(-w\in \mathcal{D}\). Moreover, for any non-unanimous profile \((w,w^{\prime },w'')\in \mathcal{D}^{3}\), one has from property M that φ(w,w′,w′′)=−φ m (w,w′,w′′). Since \(\mathcal{D}\) is 3-stable under φ m , then \(\mathcal{D}\) is 3-stable under φ. Now consider profile \(\widetilde{\pi }=(x,y,z,t)\). Observe that ∀d, \(n_{a}(\widetilde{\pi }\mid _{d})=3\) and \(n_{-a}(\widetilde{\pi }\mid _{d})=1\). From independence together with anonymity, we get that \(\varphi (\widetilde{\pi })=\{(a,a,a,a), (-a,-a,-a,-a)\}\). Thus \(\varphi (\widetilde{\pi })\notin \mathcal{D}\), therefore φ is not triple-consistent, a contradiction. Finally, independence together with anonymity ensures that \(\forall \overline{\mathcal{D}} \in \varDelta \), \(\forall \overline{\pi }\in \overline{\mathcal{D}}_{\varphi }^{3}\), \(n_{a}(\overline{\pi }\mid _{d})=3\Rightarrow [ \varphi (\overline{\pi })]^{d}=-a\), and the proof is complete. □

Lemma 7

Let φ be triple-consistent and such that property M holds. Let \(h>h^{\prime }\in \mathbb{N} \) with h+hodd. Consider any domain \(\mathcal{D}\). Then for any \(\pi \in \mathcal{D}_{\varphi }^{h+h^{\prime }}\), for any a∈{−1,1} and for any issue d,

  1. (1)

    if [n a (π d )=h and n a (π d )=h′⇒[φ(π)]d=−a], then [n a (π d )=n⇒[φ(π)]d=a];

  2. (2)

    if [n a (π d )=h and n a (π d )=h′⇒[φ(π)]d=a], then [n a (π d )=n⇒[φ(π)]d=−a].

Proof

From Lemma 6, \(\varphi =\varphi _{m}^{-}\) in restriction to 3-program profiles. Define \(\mathcal{D}=\{x,y,z,-x,-y,-z\}\subset \{-1,1\}^{3}\) by x=(−a,a,a), y=(−a,a,−a) and z=(−a,−a,a). Using property M , it is easily checked that \(\mathcal{D}\) is 3-stable under φ. Consider the \((\mathcal{D},h+h^{\prime })\)-profile \(\pi =(\overset{h-h^{\prime }}{\overbrace{x,\ldots,x}},\overset{h^{\prime }}{\overbrace{y,\ldots,y}},\overset{h^{\prime }}{\overbrace{z,\ldots,z}})\). Then n a (π d )=h and n a (π d )=h′ for d=2,3. Suppose that [φ(π)]2=−a. From independence and anonymity, we get [φ(π)]3=−a. Moreover, triple-consistency requires φ(π)=(a,−a,−a). Since n a (π1)=h+h′, then we must have [φ(π)]1=a, and assertion (1) follows from independence and anonymity. Similarly, if [φ(π)]2=a, then [φ(π)]3=a, while triple-consistency requires φ(π)=(−a,a,a). Since n a (π1)=h+h′, then we must have [φ(π)]1=−a, and assertion (2) follows from independence and anonymity. □

Lemma 8

Let φ be triple-consistent and such that property M holds. For any domain \(\mathcal{D}\), for any \(n\in \mathbb{N} \) and any \(\pi \in \mathcal{D}_{\varphi _{m}}^{n}\), \(\varphi (\pi )\in \{\varphi _{m}(\pi ), \varphi _{m}^{-}(\pi )\}\).

Proof

Consider any domain \(\mathcal{D}\). Suppose first that there exists a∈{1,−1} together with \(\{h,h^{\prime },k,k^{\prime }\}\subset \mathbb{N} \) with h>h′, k>k′, and n=h+h′=k+k′, such that, for two \((\mathcal{D},n)\)-profiles \(\pi ,\pi ^{\prime }\in \mathcal{D}_{\varphi }^{n}\), one has for some issue d:

  1. (1)

    n a (π d )=h, n a (π d )=h′, and [φ(π)]d=a;

  2. (2)

    n a (π′∣ d )=k, n a (π′∣ d )=k′, and [φ(′π)]d=−a.

It follows from Lemma 7 that for any issue d′, [n a (π d)=n⇒[φ(π)]d=−a] and [n a (π′∣ d)=n⇒[φ(π)]d=a], which clearly contradicts hat φ is independent and anonymous. Thus, we have proven that:

  1. (3)

    a∈{1,−1}, ∀d, either [n a (π d )>n a (π d ) and [φ(π)]d=a], or [n a (π d )>n a (π d ) and [φ(π)]d=−a].

Next, suppose that there exist \(\mathcal{D}\in \varDelta \), \(n \in \mathbb{N} \) and a∈{1,−1} such that for some \(\pi \in \mathcal{D}_{\varphi }^{n}\) and some d, we have

  1. (4)

    [n a (π d )>n a (π d )⇒[φ(π)]d=−a] and [n a (π d )>n a (π d )⇒[φ(π)]d=−a].

Consider \(\mathcal{D}^{\ast }=\{x,y\}\subset \{-1,1\}^{2}\), with x=(a,−a) and y=(−a,a). It follows from Lemma 6 that φ(x,x,y)=y and φ(x,y,y)=x, and thus \(\mathcal{D}^{\ast }\) is 3-stable under φ. Then define the \((\mathcal{D}^{\ast },t+t^{\prime })\)-profile \(\pi =(\overset{t}{\overbrace{x,\ldots,x}},\overset{t^{\prime }}{\overbrace{y,\ldots,y})}\), where t>t′. One gets from independence together with (4) that \(\varphi (\pi )=(-a,-a)\notin \mathcal{D}^{\ast }\), therefore φ is not triple-consistent, a contradiction. Thus, we have either [n a (π d )>n a (π d )⇒[φ(π)]d=−a] or [n a (π d )>n a (π d )⇒[φ(π)]d=−a]. Using anonymity together with independence, this proves that for all \(\mathcal{D}\in \varDelta \), for all d, for all \(n \in \mathbb{N} \) and all a∈{1,−1}, and for all \(\pi \in \mathcal{D}_{\varphi _{m}}^{n}\), either

  1. (5)

    [n a (π d )>n a (π d )⇒[φ(π)]d=a] and [n a (π d )>n a (π d )⇒[φ(π)]d=−a]

or

  1. (6)

    [n a (π d )>n a (π d )⇒[φ(π)]d=−a] and [n a (π d )>n a (π d )⇒[φ(π)]d=a].

Since (5) defines φ m while (6) defines \(\varphi _{m}^{-}\), the proof is complete. □

The proof of Theorem 3 is now complete. Indeed, consider any anonymous, independent and triple-consistent aggregation rule φ. If M + holds, then Lemma 4 implies that φ=φ m . If M = holds, then, from Lemma 5, φ is almost constant. If M holds, Lemma 8 imply that \(\varphi \in \varPhi _{m}^{+-}\).

Rights and permissions

Reprints and permissions

About this article

Cite this article

Laffond, G., Lainé, J. Triple-consistent social choice and the majority rule. TOP 22, 784–799 (2014). https://doi.org/10.1007/s11750-013-0300-1

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s11750-013-0300-1

Keywords

Mathematics Subject Classification (2010)

Navigation