2009/2/4 産業組織研究会 matching as signal kyushu university nobuaki hori
TRANSCRIPT
04/21/23
産業組織研究会
Matching as SignalMatching as Signal
Kyushu UniversityKyushu University
Nobuaki HoriNobuaki Hori
2
Introduction
• Motivation– Why is ranking of universities (vertically
and/or horizontally) stable, (in some cases,) irrelevantly to the quality of research or instruction ?
• ⇒Matching + Signaling– Matching between workers and schools– Names of universities work as “signals”
3
Introduction
• Signaling with a continuum of differential types– Mailath (1987): direct choice of vertical signal
(education) – This model : matching → Pooling is robust
• Matching between the workers and univ.– Epple, Romano and Sieg (2006), Akabayashi a
nd Naoi (2008)↑Univ. are intrinsically differentiated.– This model: potentially identical
4
Introduction
• Matching tournament– Cole, Mailath and Postlewaite (1992, 95, 98,
2001), Hopkins(2006) : matching is goal.– This model: matching is instrument.
5
Main results
• Various types of ranking are realized as a self-enforcing belief
• While potentially identical, highly ranked universities can enjoy the status by charging higher tuition fees.
• When matching is assortative, a rent absorption of tuition fee implements efficient level of educational performance.
6
ModelModel
• Workers (=students)Workers (=students)– heterogeneous in innate abilities, heterogeneous in innate abilities, aa
• Educational performance, Educational performance, ee– workplace productivityworkplace productivity– cost of performancecost of performance
• AssumptionsAssumptions
min max[ , ] ~ ( ), ( ) ia a a F a F a i
( , )iy e a
( , )ic e a
2( , ) ( , ) ( , )0, 0, 0i i iy e a y e a y e a
e a e a
2( , ) ( , ) ( , )0, 0, 0i i ic e a c e a c e a
e a e a
7
Model
• Utility functionUtility function
• • But But ee and and aa are not observable are not observable
⇒ ⇒ Incentives for Incentives for SignalingSignaling• Matches between workers and Matches between workers and
universities work as a signaling universities work as a signaling devicedevice.
, (productivity) (wage)e a y w
( , )
(wage) (cost of effort) (tuition fee)
u w c e a p
8
Timing
1. Universities simultaneously set tuition fees p(j)
↓↓
2. Students and universities make one-to-one matches, bargaining over educational performances e
↓↓
3. Workers are paid w(s) in the labor market.
9
ModelModel
• UniversitiesUniversities– Continuous, indexed byContinuous, indexed by– ⇒ ⇒ Positive measure of students can Positive measure of students can
not go to universitiesnot go to universities
• Lexicographic Preference (1)p (2)e
• Interpretation of educational performanceInterpretation of educational performance1.1. educational service offered by universitieseducational service offered by universities
2.2. score of achievement test prior to score of achievement test prior to entranceentrance
0 0[ ,1], 0 1 j J
00
10
Model
• Assumptions for the labor marketAssumptions for the labor market– A worker’s info.A worker’s info.
– Perfect competitionPerfect competition
(educational performance) not observable
(innate ability) not observable
(educational t observablrack) e
e
a
s
0[ ,1] s J
( ) [ | ]w s E y s
11
Equilibrium ConceptsEquilibrium Concepts
1.1. Perfect Bayesian Nash eq.Perfect Bayesian Nash eq.
2.2. Stable matchingStable matching
* * *( ( )) ( ( )) ( ( ( )), )
( ) ( ) ( ( ), ) for any i i i i
i
w s a p s a c e s a a
w s p s c e s a s J
*( ) : equilibrium of
( ) : necessary to get record i is a s a
e s e s
12
Image of self-enforcing rank
• Universities are ranked by Universities are ranked by – ⇒ ⇒ Matching is “assortative”Matching is “assortative”
( )w j
( )w j ( )e j
( )a j
13
A social belief
• J (set of university) could have some pooling sub-ranges,
0 1 1 2 1[ , ),[ , ), [ ,1), 1, 2, , M M
0 1 1where 1 M
14
A social belief
• : The probability that the innate ability of the graduate is higher than that of . – When the both belong to a same sub-
range
– Otherwise
Pr[ ( ) ( )]a j a j j
j
Pr[ ( ) ( )] Pr[ ( ) ( )] 1/ 2a j a j a j a j
, Pr[ ( ) ( )] 1j j a j a j For
15
The social belief (Image M=2)
z
1
1
1l
3l
4l
2l
1l 2l 3l 4l(agents)i
(universities)j
z
16
Matching stage
• is given.• stable matching
* * *( ( )) ( ( )) ( ( ( )), )
( ) ( ) ( ( ), ) for any i i i i
i
w s a p s a c e s a a
w s p s c e s a s J
( )p j
* equilibrium of ( ) :
necessary to get record ( ) : ii s as a
e se s
17
Matching stage
Lemma 1Lemma 1 If an equilibrium contains a If an equilibrium contains a pooling range for any pair pooling range for any pair universities which belong to it,universities which belong to it,
• : equilibrium performance of : equilibrium performance of aa• : equilibrium payment of : equilibrium payment of aa
Lemma 2 Lemma 2 In equilibrium,In equilibrium,
( ) ( ), ( ) ( ), ( ) ( )p j p j e j e j w j w j
( )a( )a
( )a a is weakly increasing in
18
Matching stage
• rational expectationsrational expectations
• In equilibrium,
1*1( ( )) ( ( ), ) ( ) (
m
m
a
i i m maw s a y a a dF a
)
*( ( )) ( )e s a a
*( ( )) ( )p s a a
19
Image of stable match (separating)
2
12 1 1( , ) ( ) / ( ) ( )
m
m
a
m m may e a F a
( , )iu s i a
( , )iu s i a ( , )iu s i a
e1( )ma( )ma1( )ma
11 1( , ) ( ) / ( ) ( )
m
m
a
m m may e a F a
20
Boundary conditions
• Initial boundary condition
• Boundary condition for two ranges
0
min
0 0 0 0 0
0
(0, ) ( )( ( ), ) ( ) ( ( ), )
a
ay a dF a
y a a a c a a
1
1
111 1
1
( ( ), ) ( )( ) ( ( ), )
( ( ), ) ( )( ) ( ( ), )
m
mmm m m
m m
m
mmkm km km
m m
y a a dF aa c a a
y a a dF aa c a a
21
Image of stable match (separating)
1
1
( , ) ( )( )
m
m
a
am
m m
y e a dF aa
1( , ) m mu s a
e( )ma1( )ma
1( , ) m mu s a
1( )ma
( )
( ), m
m
a
a m m
recursively determi is
from
ned
22
Fee setting stage
Definition
Lemma 3
1
1
ˆIf ( ) ( , ) for any ,
ˆa same tuition fee which implements ( ) ( , )
can be best responces for the universities on the range .
m m m
m m m
a e m m
a e
m
1
1
1
ˆ( , )
arg max ( , ) ( ) / ( , )
m m
m
m m mm
e
y e a dF a c e a
23
Graphical image
1
1( ( ), ) ( ) /
m
m m m mmEy y a a dF a
1( )ma 1ˆ( , )m me
*( )ma
1
1
( )
( )
m
m
w
a
( )m mEy a
e
24
Equilibrium performances
1
Proposition 1
ˆ( ) ( , ) can be
a perfect Bayesian Nash equilibrium.=m m ma e
0 1 min 0ˆ( ) 0 ( , ), [ , ) a e a a a In equilibrium,
for
0 0 1 1 2ˆ ˆ( ) ( ) ( ) a e e
( ) 0ma m recursively determine for are
by adopting
d
Lemma
25
Equilibrium performances
0amina maxa1a 2a 4a3a
( )a*( )e a
a
e
26
Tuition fees
1
0
0
min
0 1
0 0 1 01 0
0
ˆ( ( , ), ) ( )ˆ( ) ( ( , ), )
(0, ) ( )
a
a
a
a
y e a dF aa c e a
y a dF a
Proposition 2
Equilibrium tuition fee schedule is given by
(initial condition)
27
1
1
1
1
11
1
11
( ) ( )
ˆ( ( , ), ) ( )ˆ( ( , ), )
ˆ( ( , ), ) ( )ˆ( ( , ), )
m
m
m
m
m m
a
m mam m m
m m
a
m mam m m
m m
a a
y e a dF ac e a
y e a dF ac e a
and
28
Equilibrium performances
0amina maxa1a 2a 4a3a
( )a
a
p