Market Design

Econ3052
NTU
Fall 2023
Problem Set 3

1. Watch the following video about JUPUS (https://www.youtube.com/watch?v=c-

d0f0A_KsE), the college major matching mechanism in Hong Kong.
Summarize the matching algorithm described in the video. Explain
through an example why the algorithm is equivalent to the Deferred
Acceptance Mechanism with student proposing.

2. Consider a marriage market where (1) all participants have strict pref-
erences over agents on the other side and (2) all women have the same
preferences over men. Show that there is a unique stable matching.

3. Consider the following matching with four men and five women whose
preferences are

m1 m2 m3 m4
w1 w2 w3 w4 w5
w1 w1 w4 w5
m4 m1 m1 m4 m4
w5 w3 w2 w3
, m2 m2 m3 m2 m2
w4 w5 w3 w2
m1 m3 m4 m1 m3
w3 w2 w5 w1
m3 m4 m2 m3 m1
w2 w4 w1 w4

(a) Find the DA outcome with women proposing.

(b) Does any man have incentives to misreport his true preferences?

4. Consider the medical resident matching problem with two hospitals

(h1, h2), one single student (s), and one couple (c1, c2). Each hospital
can hire one student. The preferences are

h1 h2 s c1,c2
c1 s , h1 (h1,h2) .
s c2 h2

Show that there is no stable matching.

1
In JUPAS ,

every students submit their preferences by ranking their preferred college majors in

order. Each college also rank students based on their academic qualifications and other

relevant . Students
criteria act as
proposers while each college has a
quota for the number

of students
they can admit. For example ,
the preferences for students and colleges are

S S2 53 Cl C2 (3 Cl C2 13

axx
SS S2 ⑤
2
# 2 CI rejected S2 52 + 13

S"
.

=
,

S3 S2 C3
S
rejected S3 , S3- (1
222 S2SIS '
C

S3 CI Si -
rejected S1 ,
22

S
93 SIS2 Final matching

Shaply's for
By Gale and Theorem ,
any preferences ,
there always exists at least one stable matching.
And since all have the preferences to the set of and
women same ,
proposing same men men

always choose the same proposals ,

the final matching will be the same regardless of the order in

which proposals made By the end of algorithm is

paired with his most preferred
are .
,
every man

who has there's

woman not been rejected by any
other man .
If a
blocking pair ,
it means

that the will the the algorithm she'll have

man have proposed to woman
during ,
and accepted

his she
proposal If .
doesn't ,
it means she had a better offer ,
contradicting the assumption
that she prefers the man over her current partner
.

il
-

(a) m
, Mz my M4

, &, M Wi ,
m.
rejects Wa ,
Wit Mu

Wa Ma rejects W.W4 ,
Wi Wat M2
.

Wi W , my
rejects Wh , Wa
,
WafMs , Nq-M
,

W4 Wz m,
rejects Ws, Wa >
-

Ms

#3 Ms
rejects Ws Ws > M4
-

s my rejects Ws ,
Wat Ma
As Ws
m2 rejects ,
Not Ws

W5
W4
matching
W. Wa Final

(b) No ,
since
submitting his true preferences is a dominant
strategy.
Since MCh1) = C1 , M(h2) =
S , M(C1 (2) , can not be satisfied at the same time . However ,
if

M(h1) = S , M1hz) = Cn ,
it can't be satisfied either. There's always blocking pair . Thus ,
there's

stable
no
matching.
 Market Design
Econ3052
NTU
Fall 2023
Problem Set 3

1. Watch the following video about JUPUS (https://www.youtube.com/watch?v=c-

d0f0A_KsE), the college major matching mechanism in Hong Kong.
Summarize the matching algorithm described in the video. Explain
through an example why the algorithm is equivalent to the Deferred
Acceptance Mechanism with student proposing.
Ans. The Jupas algorithm consists of two stages. During the first
stage, each program tentatively fills its slots with the best students
who apply to it as the first choice. A temporary cuto§ equal to the
lowest position among the tentatively accepted students is set for each
program. Any students strictly below the cuto§ of a program will
be rejected by the program. Let qk denote the number of open slots
of program k. At the end of first stage, any tentatively accepted
students who are ranked at the top qk of the order list will be assigned
to the program and removed from the algorithm. The number of open
slots will be adjusted accordingly. The process will now proceed to
the second stage. During the second stage, all currently unmatched
students will be processed one by one according to some chosen order.
When a student’s turn comes up, the algorithm will check whether
this student is currently tentatively accepted by her first choice. If so,
the algorithm will move to the next student. If a student has been
rejected in the last round, then the algorithm will check whether her
position at her top choice among programs that have not rejected her
is higher than or equal to the number of currently open slots. If no, the
algorithm will move to the next student. If yes, she will be assigned to
that program and removed from the algorithm. The number of open
slots will be reduced by one, and the last tentatively accepted student
is rejected. The program’s cuto§ will be moved up to the position of
the last first-choice applicants who are still tentatively accepted. The
algorithm will check whether any tentatively accepted students at their
current top choice will now rank at a position higher than the number
of remaining slots. All such students will be accepted by the program
and removed as in the first stage. After adjusting for the quotas, the
algorithm will move to the next student. The process continue until
either all students are assigned or rejected by all schools.

1
The Jupas algorithm is the same as DS with student proposing. In-
tuitively, under both algorithms, a student will assigned to her top
choice unless more higher ranked students also prefer that program.
However, a formal proof is not easy. But we can show that they pro-
duce the same result in an example. Consider the matching problem
in Q3. Suppose we run the Jupas algorithm. At the first stage, the
outcome is as followed.
m1 m2 m3 m4
w1 w2 w3 w4 w5
w1 w1 w4 w5
m4 m1 m1 m4 m4
w5 w3 w2 w3
m2 m2 m3 m2 m2
w4 w5 w3 w2
m1 m3 m4 m1 m3
w3  w2 w5 w1
m3 m4 m2 m3 m1
w2  w4 w1 w4 
m4 is matched to w5 as the latter is the top choice of m4 . After the
assigned agents and rejected applications are removed, what remains
are
m1 m2 m3
w1 w2 w3 w4
w1 w1 w4
w3 w2
m2 m2 m3 m2
w4 w3
m1 m3 m1
w2
m3 m2 m3
w4 w1
Consider w1 . Her current top choice is m2 . Since m2 ranks w1 the
highest, m2 is matched to w1 . After removing all the assigned agents,
the following remains:
m1 m3
w2 w3 w4
w4
w2
m3
w4 w3
m3 m1
m3

At this point, w4 can be assigned to m1 as it ranks at the top of m1 .

So what remains are
m3 w2 w3

w2 m3
w3 m3

2
 So finally w2 is assigned to m3 . The final match is identical to DA
with student proposing.

2. Consider a marriage market where (1) all participants have strict pref-
erences over agents on the other side and (2) all women have the same
preferences over men. Show that there is a unique stable matching.
Ans. Suppose all women have preferences: m1 , m2 , ..., mn . In any
stable matching, m1 must be matched to his favorite woman, say w1 .
(If not, (m1 , w1 ) will form a blocking pair.) Similarly, m2 must be
matched to his favorite woman among the set of all women minus
w1 . More generally, every man mi must be matched to his favorite
women among the women not matched to m1 , ..., mi1 . Intuitively, in
this case the unique matching is equivalent to matching formed by a
serial dictatorship where more popular men choose earlier. The point
is that multiple stable matching exist only when participants have
heterogeneous preferences.

3. Consider the following matching with four men and five women whose
preferences are

m1 m2 m3 m4
w1 w2 w3 w4 w5
w1 w1 w4 w5
m4 m1 m1 m4 m4
w5 w3 w2 w3
, m2 m2 m3 m2 m2
w4 w5 w3 w2
m1 m3 m4 m1 m3
w3 w2 w5 w1
m3 m4 m2 m3 m1
w2 w4 w1 w4

(a) Find the DA outcome with women proposing.

Ans:
m1 m2 m3 m4
w2 , w3 w1 , w4 , [w5 ]
w2 , [w1 ], w4
[w4 ] [w2 ]
w3
w3
w3
Final match:
m1 m2 m3 m4
w4 w1 w2 w5

3
 (b) Does any man have incentives to misreport his true preferences?
Ans: No. Note that w5 and m4 like each other the most. So
obviously, they have no incentives to misreport. Furthermore,
they will be matched to each other no matter how others report.
Other than m4 , w1 likes m2 the most, and m2 likes w1 the most.
So m2 will not again from misreporting. Given that w1 and w5
are not reachable by m1 , w4 is already the best m1 can obtained
no matter how he reports. Similarly, without w4 , w2 is the best
that m3 can hope for.

4. Consider the medical resident matching problem with two hospitals

(h1, h2), one single student (s), and one couple (c1, c2). Each hospital
can hire one student. The preferences are

h1 h2 s c1,c2
c1 s , h1 (h1,h2) .
s c2 h2

Show that there is no stable matching.

Ans. The couple only accepts (h1,h2). So either they are matched to
(h1,h2) or they are unmatched. The match (h1,c1),(h2,c2) is unstable
as it will be blocked by (h2 s). Supposed (c1, c2) are unmatched.
Matching s to h1 will be blocked by (h1,c1),(h2,c2). Matching s1 to
h2 will be blocked by (h1,s).