Abstract
In this paper, we characterize the strongly stable fractional matchings for the marriage model as the union of the convex hull of connected sets of stable matchings. Moreover, we present an algorithm that computes the set of matchings necessary to generate the above-mentioned connected sets. Finally, we show that the set of strongly stable fractional matchings has a lattice structure.
Similar content being viewed by others
Notes
For more details, see Roth and Sotomayor (1992).
Any (non-integer) solution of the above-mentioned system of linear inequalities must be a convex combination of permutation matrices.
Theorem (Roth et al. (1993)): The integer solutions of (LP) are exactly the stable matchings.
For more details, see Gusfield and Irving’s book (Gusfield and Irving 1989). The cycles are called rotations.
Notice that if \(K=\emptyset \), then \(\mu [K]=\mu \).
Using the Complementary Slackness Theorem for \(\left( LP\right) \), in Lemma 9, Roth et al. (1993) prove that when \(x_{m,w}>0\) for some stable fractional matching x, then for each stable fractional matching \(x^{\prime }\), \(\sum _{j>_{m}w}x_{m,j}^{\prime }+\sum _{i>_{w}m}x_{i,w}^{\prime }+x_{m,w}^{\prime }=1\). Here, we have that \({\bar{x}}\) is a stable fractional matching in the matching model \((M,W,P^{\mu })\).
Here, we use “\(\subset \)” to denote the strict inclusion; that is to say, \(A\subset B\) means that A is a proper subset of B.
For ease of presentation, we will denote for instance \(\mu _{12}=\mu _{1}[\sigma _{2}]=\mu [\sigma _{1},\sigma _{2}]\). Notice that by Lemma 4, we have that \(\mu _{12}=\mu [\sigma _{1},\sigma _{2}]=\mu [\sigma _{2},\sigma _{1}]=\mu _{21}\).
Lemma 4 ensures that \(\mu [\sigma _{1},\sigma _{2}]=\mu [\sigma _{2},\sigma _{1}]\).
Theorem (Conway)When preferences are strict, the set of stable matchings is a distributive lattice under common order of men, dual to the common order of women.
For more details on lattice theory, see Birkhoff (1948).
The partial orders \(\succeq _{M}\) and \(\succeq _{W}\) are formally presented in the Appendix.
References
Birkhoff G (1946) Tres observaciones sobre el algebra lineal. Univ Nac Tucumán Rev Ser A 5:147–151
Birkhoff G (1948) Lattice theory, vol 25. American Mathematical Society, New York
Gale D, Shapley L (1962) College admissions and the stability of marriage. Am Math Monthly 69(1):9–15
Gusfield D, Irving R (1989) The stable marriage problem: structure and algorithms. MIT Press, Cambridge
Irving R, Leather P (1986) The complexity of counting stable marriages. SIAM J Comput 15(3):655–667
Knuth D (1997) Stable marriage and its relation to other combinatorial problems: an introduction to the mathematical analysis of algorithms, vol 10. American Mathematical Soc, New York
McVitie D, Wilson L (1970) Stable marriage assignment for unequal sets. BIT Numer Math 10(3):295–309
Roth A, Sotomayor M (1992) Two-sided matching: a study in game-theoretic modeling and analysis. Cambidge University Press, Cambridge
Roth A, Rothblum U, Vande Vate J (1993) Stable matchings, optimal assignments, and linear programming. Math Oper Res 18(4):803–828
Rothblum U (1992) Characterization of stable matchings as extreme points of a polytope. Math Program 54(1–3):57–67
Shapley L, Shubik M (1972) The assignment game I: the core. Int Game Theory 1
Vande Vate J (1989) Linear programming brings marital bliss. Oper Res Lett 8(3):147–153
Acknowledgements
We are grateful to Jordi Massó, and the members of the Game Theory Group of IMASL for their queries and comments. In addition, we are grateful to the members of GAECI, especially Graciela Lucero Arrúa, for their assistance with the English writing. Our work is partially supported by the Universidad Nacional de San Luis, through Grant 319502, and by the Consejo Nacional de Investigaciones Científicas y Técnicas (CONICET), through Grant PIP 112-200801-00655. We are grateful to the anonymous referees for their valuable comments.
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendix
Appendix
Given two stable fractional matchings x and y, we say that x weakly dominates y in man \(m^{\prime }s\) opinion, if
for all w, which is denoted by \(x\succeq _{m}y\). We also say that x strongly dominates y, \(\left( x\succ _{m}y\right) \), if at least one inequality is strict. We can define domination in a woman’s opinion analogously.
Also, we say that x weakly dominates y in man \(m^{\prime }s\) opinion under a profile of reduced lists \(P^{\mu }\), if
for all \(w\in {P^{\mu }(m)}\), which is denoted by \(x\succeq ^{\mu }_{m}y\). We also say that x strongly dominates y, \(\left( x\succ ^{\mu } _{m}y\right) \), if at least one inequality is strict. We can define domination in a woman’s opinion under a profile of reduced lists, in the same way.
A partial order \(\succeq _{M}\) can be defined on the set of stable fractional matchings where \(x\succeq _{M}y\), if \(x\succeq _{m}y\) for each \(m\in M\). That is, \(x\succeq _{M}y\), meaning that x weakly dominates y in every man’s opinion. This also applies for \(\succeq ^{\mu }_{M}\) in the reduced preference lists \(P^{\mu }\). In the same way, we can define domination in every woman’s opinion (\(\succeq _{W}\) and \(\succeq ^{\mu }_{W}\)).
Remark 3
Let x be a stable fractional matching in the matching model \(\left( M,W,P\right) \). Since x is a convex combination of stable matchings, then \(x^{\mu _{M}}\succeq _{M}x\succeq _{M}x^{\mu _{W}}\) and \(x^{\mu _{W}}\succeq _{W}x\succeq _{W}x^{\mu _{M}}\). Moreover, if x is a stable fractional matching for the matching model \(\left( M,W,P^{\mu }\right) \), then \(x^{\mu }\succeq _{M}x\succeq _{M}x^{\mu _{W}}\) and \(x^{\mu _{W}}\succeq _{W}x\succeq _{W}x^{\mu }\). This follows from Lemma 1, which means that each stable matching for the matching model \(\left( M,W,P^{\mu }\right) \) is also a stable matching for the matching model \(\left( M,W,P\right) \).
Lemma 6
Let (M, W, P) be a matching model. Let \(\mu \in S\left( P\right) \), and \(P^{\mu }\) be the reduced lists. Let \({\bar{x}}\) be a stable fractional matching in the matching model (M, W, P). If \({\bar{x}}\) is strongly stable for the matching model \(\left( M,W,P^{\mu }\right) \), then \({\bar{x}}\) is strongly stable for the matching model \(\left( M,W,P\right) \).
Proof
Let \(\left( M,W,P\right) \) be a matching model. Let \(\mu \in S\left( P\right) \), and \(P^{\mu }\) be the reduced lists. Let \({\bar{x}}\) be a stable fractional matching in the matching model (M, W, P) and let us assume that \({\bar{x}}\) is strongly stable in \(\left( M,W,P^{\mu }\right) \). Hence:
holds for each \(\left( m,w\right) \in A\left( P^{\mu }\right) \). Then, we will prove that
holds for each \(\left( m,w\right) \in A\left( P\right) \), as well.
We will consider the following cases:
- (i)
Let \(\left( m,w\right) \in A\left( P^{\mu }\right) \). That is, \(\left( m,w\right) \) was not eliminated in \(P^{\mu }\). Therefore
$$\begin{aligned} \sum _{j\ge _{m}w}{\bar{x}}_{m,j}\ge \sum _{j\ge _{m}^{\mu }w}{\bar{x}}_{m,j} \end{aligned}$$holds, since for each man m, there are more women in his original list of preferences than in his reduced list of preferences.
Hence,
$$\begin{aligned} 1-\sum _{j\ge _{m}w}{\bar{x}}_{m,j}\le 1-\sum _{j\ge _{m}^{\mu }w}{\bar{x}}_{m,j}. \end{aligned}$$With a similar argument, we have the following:
$$\begin{aligned} 1-\sum _{i\ge _{w}m}{\bar{x}}_{i,w}\le 1-\sum _{i\ge _{w}^{\mu }m}{\bar{x}}_{i,w}. \end{aligned}$$By hypothesis and constraints (1) and (2) in (LP):
$$\begin{aligned} 0= & {} \left[ 1-\sum _{j\ge _{m}^{\mu }w}{\bar{x}}_{m,j}\right] \cdot \left[ 1-\sum _{i\ge _{w}^{\mu }m}{\bar{x}}_{i,w}\right] \\\ge & {} \left[ 1-\sum _{j\ge _{m}w} {\bar{x}}_{m,j}\right] \cdot \left[ 1-\sum _{i\ge _{w}m}{\bar{x}}_{i,w}\right] \ge 0. \end{aligned}$$Then, for \(\left( m,w\right) \in A\left( P^{\mu }\right) \)
$$\begin{aligned} \left[ 1-\sum _{j\ge _{m}w}{\bar{x}}_{m,j}\right] \cdot \left[ 1-\sum _{i\ge _{w} m}{\bar{x}}_{i,w}\right] =0. \end{aligned}$$ - (ii)
Let \(\left( m,w\right) \in A\left( P\right) -A\left( P^{\mu }\right) \). Considering that \(\left( m,w\right) \) was eliminated in \(P^{\mu }\), we analyze three cases separately:
- (a)
If \(w>_{m}\mu (m)\).
Since \(x^{\mu }\succeq _{M}{\bar{x}}\), then
$$\begin{aligned} \sum _{j>_{m}w}{\bar{x}}_{m,j}\le \sum _{j>_{m}w}x_{m,j}^{\mu }=0. \end{aligned}$$(14)As \({\bar{x}}\) is a stable fractional matching in (M, W, P), constraints (2) and (3) of (LP) hold, and by (14), we have \(\sum _{i\ge _{w} m}{\bar{x}}_{i,w}=1\). Then:
$$\begin{aligned} \left[ 1-\sum _{j\ge _{m}w}{\bar{x}}_{m,j}\right] \cdot \left[ 1-\sum _{i\ge _{w} m}{\bar{x}}_{i,w}\right] =0. \end{aligned}$$ - (b)
If \(w<_{m}\mu _{W}(m)\).
Since \({\bar{x}}\) is a stable fractional matching, we have that \({\bar{x}}\) is a convex combination of stable matchings. Then, \({\bar{x}}\succeq _{M}x^{\mu _{W}}\) holds by Remark 3. Therefore, we also have that \(\sum _{j\ge _{m}w}{\bar{x}}_{m,j} \ge \sum _{j\ge _{m}w}x_{m,j}^{\mu _{W}}=1\). Then:
$$\begin{aligned} \left[ 1-\sum _{j\ge _{m}w}{\bar{x}}_{m,j}\right] \cdot \left[ 1-\sum _{i\ge _{w} m}{\bar{x}}_{i,w}\right] =0. \end{aligned}$$ - (c)
If \(\mu (m)>_{m}w>_{m}\mu _{W}(m)\).
Since \(\left( m,w\right) \not \in A\left( P^{\mu }\right) \), man m was eliminated by woman w in \(P^{\mu }\). If \(m>_{w}\mu (w)\), then (m, w) is a blocking pair for the stable matching \(\mu _{W}\), and this is a. Hence, \(\mu (w)>_{w}m\).
Since \({\bar{x}}\succeq _{W}x^{\mu }\) by Remark 3, we have \(\sum _{i\ge _{w}m}{\bar{x}}_{i,w}\ge \sum _{i\ge _{w}m}x_{i,w}^{\mu }=x^{\mu }_{\mu (w),w}=1\). Using constraint \(\left( 2\right) \) of (LP) we have that
$$\begin{aligned} \sum _{i\ge _{w}m}{\bar{x}}_{i,w}=1; \end{aligned}$$then
$$\begin{aligned} \left[ 1-\sum _{j\ge _{m}w}{\bar{x}}_{m,j}\right] \cdot \left[ 1-\sum _{i\ge _{w} m}{\bar{x}}_{i,w}\right] =0. \end{aligned}$$
- (a)
From cases (i) and (ii), we conclude that \({\bar{x}}\) is strongly stable in \(\left( M,W,P\right) \). \(\square \)
Proof of Claim
We need to prove that there is a pair \(({\bar{m}},{\bar{w}})\in {A(P^{\mu _{{\bar{x}}}})}\), such that \(\mu _{{\bar{x}}}({\bar{m}})>_{{\bar{m}}}{\bar{w}}>_{{\bar{m}}}\mu ^{l}({\bar{m}})\) and \(\mu ^{l}({\bar{w}})>_{{\bar{w}}}{\bar{m}}>_{{\bar{w}}}\mu _{{\bar{x}}}({\bar{w}})\).
To prove \(\mu _{{\bar{x}}}({\bar{m}})>_{{\bar{m}}}{\bar{w}}>_{{\bar{m}}}\mu ^{l}({\bar{m}})\), we will analyze two cases:
- (i)
Assume that there exists \({\bar{\sigma }}\in {{\varPhi }(\mu _{{\bar{x}}})}\) and \({\bar{m}}\in {{\bar{\sigma }}}\), such that \(\mu _{{\bar{x}}}[{\bar{\sigma }}]({\bar{m}})>_{{\bar{m}}}\mu ^{l}({\bar{m}})\). Let \({\bar{w}}=\mu _{{\bar{x}}}[{\bar{\sigma }}]({\bar{m}})\). Then, \(({\bar{m}},{\bar{w}})\in {A(P^{\mu _{{\bar{x}}}})}\). Since \({\bar{m}}\in {{\bar{\sigma }}}\), we have that \(\mu _{{\bar{x}}}({\bar{m}})>_{{\bar{m}}}\mu _{{\bar{x}}}[{\bar{\sigma }}]({\bar{m}})\). Hence, for \(({\bar{m}},{\bar{w}})\in {A(P^{\mu _{{\bar{x}}}})}\), we have that \(\mu _{{\bar{x}}}({\bar{m}})>_{{\bar{m}}}{\bar{w}}>_{{\bar{m}}}\mu ^{l}({\bar{m}})\).
- (ii)
If not, for all \(\sigma \in {{\varPhi }(\mu _{{\bar{x}}})}\) and for all man \(m\in {{\bar{\sigma }}}\), we have that \(\mu _{{\bar{x}}}[{\bar{\sigma }}](m)=\mu ^{l}(m)\). Let \(M_{0}=\{m\in {M}:m\in {{\bar{\sigma }}} \text{ for } \text{ some } {\bar{\sigma }}\in {{\varPhi }(\mu _{{\bar{x}}})}\}\). Then, there exists a pair \(({\bar{m}},{\bar{w}})\in {A(P^{\mu _{{\bar{x}}}})}\), such that \({\bar{m}}\in {M\setminus M_{0}}\) and \(\mu _{{\bar{x}}}({\bar{m}})>_{{\bar{m}}}{\bar{w}}>_{{\bar{m}}}\mu ^{l}({\bar{m}})\). If not, for each \(m'\in {M\setminus M_{0}}\), there is no acceptable woman \(w'\) such that \(\mu _{{\bar{x}}}(m')>_{m'}w'>_{m'}\mu ^{l}(m')\). That is, for each \(m'\in {M\setminus M_{0}}\), either we have that \(\mu _{{\bar{x}}}(m')=\mu ^{l}(m')\), or \(\mu ^{l}(m')\) is the second woman in the reduced preference list of man \(m'\). Then \(\mu ^{l}\in {{\mathcal {M}}_{\mu _{{\bar{x}}}}}\), which results in a contradiction. Then, we have that there exists a pair \(({\bar{m}},{\bar{w}})\in {A(P^{\mu _{{\bar{x}}}})}\), such that \({\bar{m}}\in {M\setminus M_{0}}\) and \(\mu _{{\bar{x}}}({\bar{m}})>_{{\bar{m}}}{\bar{w}}>_{{\bar{m}}}\mu ^{l}({\bar{m}})\).
Now, we will prove \(\mu ^{l}({\bar{w}})>_{{\bar{w}}}{\bar{m}}>_{{\bar{w}}}\mu _{{\bar{x}}}({\bar{w}})\). By the stability of \(\mu ^{l}\), we have that \(\mu ^{l}({\bar{w}})>_{{\bar{w}}}{\bar{m}}\). We only need to prove that \({\bar{m}}>_{{\bar{w}}}\mu _{{\bar{x}}}({\bar{w}})\). If \(\mu _{{\bar{x}}}({\bar{w}})>_{{\bar{w}}}{\bar{m}}\), then in step 2 of the reduction procedure, \({\bar{m}}\) will be eliminated by woman \({\bar{w}}\) from her reduced preference list, \(P^{\mu _{{\bar{x}}}}({\bar{w}})\). Hence, in step 3, the man \({\bar{m}}\) will eliminate woman \({\bar{w}}\) from his reduced preference list, which results in a contradiction of the existence of the pair \(({\bar{m}},{\bar{w}})\in A(P^{\mu _{{\bar{x}}}})\). Then, \(\mu ^{l}({\bar{w}})>_{{\bar{w}}}{\bar{m}}>_{{\bar{w}}}\mu _{{\bar{x}}}({\bar{w}})\). \(\square \)
Rights and permissions
About this article
Cite this article
Neme, P.A., Oviedo, J. A characterization of strongly stable fractional matchings. TOP 28, 97–122 (2020). https://doi.org/10.1007/s11750-019-00528-y
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11750-019-00528-y
Keywords
- Matching markets
- Marriage model
- Stable fractional matchings
- Strongly stable fractional matchings
- Linear programming