Design and Analysis of Algorithms

Stable matching

Elias Koutsoupias
Oxford, Hilary Term 2025
Stable matching [KT 1.1]

• Stable matching is a simple game-theoretic algorithmic problem

• Multiple applications
• Nobel Prize to Lloyd Shapley and Alvin Roth, 2012

1
Matching med-school students to hospitals – overview

Goal. Given a set of preferences among hospitals and med-school

students, design a self-reinforcing admissions process.

Unstable pair. Hospital h and student s form an unstable pair if

both:

• h prefers s to one of its admitted students

• s prefers h to assigned hospital.

Stable assignment. Assignment with no unstable pairs.

• Natural and desirable condition.

• Individual self-interest prevents any hospital–student side deal.

2
Stable matching problem: input

Input: A set of n hospitals H and a set of n students S.

• Each hospital h ∈ H ranks students

• Each student s ∈ S ranks hospitals

More concretely, the input is a set of 2n permutations of (1, . . . , n).

3
Example : input

n=3
Hospitals = {A, B, C}
Students = {X, Y, Z}
Preferences of hospitals (left) and students (right)

1 2 3 1 2 3
A X Y Z X B A C
B Y X Z Y A B C
C X Y Z Z A B C

4
Perfect matching

Definition
A matching M is a set of ordered pairs h − s with h ∈ H and s ∈ S,
such that

• Each hospital h ∈ H appears in at most one pair of M

• Each student s ∈ S appears in at most one pair of M

A matching M is perfect if every member of H (and S) is matched, i.e.,

appears in M.

5
Example : perfect matching

n=3
Hospitals = {A, B, C}
Students = {X, Y, Z}
Preferences of hospitals (left) and students (right)

1 2 3 1 2 3
A X Y Z X B A C
B Y X Z Y A B C
C X Y Z Z A B C

A perfect matching: {A − Z, B − Y, C − X}.

6
Unstable pair

Definition
Given a perfect matching M, hospital h and student s form an
unstable pair if both:

• h prefers s to matched student

• s prefers h to matched hospital.

An unstable pair h − s could each improve by joint action.

7
Example : unstable pair

n=3
Hospitals = {A, B, C}
Students = {X, Y, Z}
Preferences of hospitals (left) and students (right)

1 2 3 1 2 3
A X Y Z X B A C
B Y X Z Y A B C
C X Y Z Z A B C

A − Y is an unstable pair for the given perfect matching.

8
Stable matching problem

Definition (Stable matching)

A stable matching is a perfect matching with no unstable pairs.
Stable matching problem.

Definition (Stable matching problem)

Given the preference lists of n hospitals and n students, find a stable
matching (if one exists).

1 2 3 1 2 3
A X Y Z X B A C
B Y X Z Y A B C
C X Y Z Z A B C

A stable matching {A − X, B − Y, C − Z}.

9
Stable roommate problem

• Do stable matchings always exist?

• Not obvious a priori.
Consider the following similar problem.
Stable roommate problem:
• 2n people; each person ranks others from 1 to 2n − 1
• Assign roommate pairs so that no unstable pairs.

1 2 3
A B C D
B C A D
C A D B
D C B A

Claim: Stable roommate matchings may not exist. 10

Example of roommate problem with no stable matching

Consider the following instance:

1 2 3
A B C D
B C A D
C A B D
D A B C

Note that A − B, C − D ⇒ B − C unstable

No perfect roommate matching is stable:

A − B, C − D ⇒ B − C is unstable
A − C, B − D ⇒ A − B is unstable
A − D, B − C ⇒ A − C is unstable 11
Gale-Shapley deferred acceptance algorithm

A natural algorithm that guarantees to find a stable matching.

Gale-Shapley(lists of preferences)
1 M=∅
2 while (some hospital h is unmatched and
hasn’t proposed to every student)
3 s = first student on h’s list to whom h has not yet proposed
4 if (s is unmatched)
5 Add h − s to matching M
6 else if (s prefers h to current partner h0 )
7 Replace h0 − s with h − s in matching M
8 else nothing (i.e., s rejects h)
9 return stable matching M.

12
Running time

• Hospitals propose to students in decreasing order of preference

• Once a student is matched, the student never becomes
unmatched; only “trades up.”

Lemma
Algorithm terminates after at most n2 iterations of While loop.

Proof.
Each time through the While loop, a hospital proposes to a new
student. Thus, there are at most n2 possible proposals.

13
Examples with n(n − 1) + 1 steps

1 2 3 1 2 3
A X Y Z X B C A
B Y X Z Y C A B
C X Y Z Z A B C

1 2 3 4 1 2 3 4
A W X Y Z W B C D A
B X Y W Z X C D A B
C Y W X Z Y D A B C
D W X Y Z Z A B C D

14
Proof of correctness

Lemma
The Gale-Shapley algorithm finds a perfect matching.

Proof.
• Suppose, for sake of contradiction, that some hospital h is
unmatched upon termination of the Gale–Shapley algorithm
• Then some student, say s, is unmatched upon termination
• So s was never proposed to, because once proposed it becomes
matched and remains matched thereafter
• But, h proposes to every student, since h ends up unmatched

15
Proof of correctness

Lemma
The matching M returned by the Gale-Shapley algorithm is stable.

Proof.
Consider any pair hs that is not in M. We show that it is not unstable.
h never proposed to s: Therefore, h prefers its student in M to s.
h proposed to s: Therefore s rejected h at some point, which means
that s ended up with a more preferred hospital.

Theorem (Gale–Shapley 1962)

The Gale–Shapley algorithm guarantees to find a stable matching for
any problem instance.

16
Multiple stable matchings

An instance may have multiple stable matchings. For example:

{A − X, B − Y, C − Z} and {A − Y, B − X, C − Z}

1 2 3 1 2 3
A X Y Z X B A C
B Y X Z Y A B C
C X Y Z Z A B C

1 2 3 1 2 3
A X Y Z X B A C
B Y X Z Y A B C
C X Y Z Z A B C
17
Valid partners

A student s is a valid partner for hospital h if there exists any stable

matching in which h and s are matched.
For example

• Both X and Y are valid partners for A.

• Both X and Y are valid partners for B.
• Z is the only valid partner for C.

1 2 3 1 2 3
A X Y Z X B A C
B Y X Z Y A B C
C X Y Z Z A B C

18
Which stable matching?

Hospital-optimal assignment: Each hospital receives best valid

partner.

• Is it a perfect matching?
• Is it stable?

Lemma
The Gale-Shapley algorithm returns the hospital-optimal assignment.
As a corollary, we get that the hospital-optimal assignment is stable.

19
Which stable matching?

Student-pessimal assignment: Each student receives worst valid

partner.
Lemma
The Gale-Shapley algorithm returns the student-pessimal
assignment.
As a corollary, we get that the student-pessimal assignment is stable.

20
Thuthfulness

Is the Gale-Shapley algorithm truthful? That is, can participants gain

by misrepresenting their preferences?

• A hospital cannot get a better solution by lying about their

preference
• But a student may gain by lying about their preferences

21