Econs 424 - Signalling Games I: Félix Muñoz-García
Econs 424 - Signalling Games I: Félix Muñoz-García
Félix Muñoz-García
There is a …rm and a worker. In this game, nature chooses the "type"
of the …rm (player 1). With probability p, the …rm is of high quality
(H) and, with probability 1 p, the …rm is of low quality (L).
The …rm chooses either to o¤er a job to the worker (O) or not to
o¤er a job (N).
If no job is o¤ered, the game ends and both parties receive 0.
If the …rm o¤ers a job, the worker either accepts (A) or rejects (R) the
o¤er.
Firm:
The worker’s e¤ort on the job brings the …rm a pro…t of 2.
If the worker rejects an o¤er of employment, then the …rm gets a
payo¤ of -1 (associated with being jilted).
Worker:
Rejecting an o¤er yields a payo¤ of 0 to the worker.
Accepting an o¤er yields the worker a payo¤ of 2 if the …rm is of high
quality, and -1 if the …rm is of low quality.
The worker does not observe the quality of the …rm directly.
A 2,2
NH 1 OH q
0,0
R
p
-1,0
2
2,-1
(1-p) A
0,0
NL 1 OL (1-q) R
-1,0
NH 1 OH q
0,0
R
p
-1,0
2
2,-1
(1-p) A
0,0
NL 1 OL (1-q) R
-1,0
Given this strategy from the …rm, the worker’s beliefs after observing
a job o¤er must be
p 0
q= =0
p 0+(1 p) 1
Intuitively, this implies that after observing O, the worker
concentrates all his beliefs on being in the lower node of the
information set, i.e., low-quality …rm.
Given these beliefs, it is optimal for the worker to reject the job o¤er,
since 0 > 1.
For future reference, we shade the branch in which the worker rejects
the …rm’s job o¤er, an action that is independent on the …rm’s type
(something that the worker cannot observe).
If the worker rejects, then the high-quality …rm prefers to not make a
job o¤er, N H , since 0 > 1, as prescribed by this strategy pro…le.
However, the low-quality …rm prefers to not make job o¤ers, N L , which
contradicts the initially proposed strategy pro…le N H O L .
Then, the separating strategy pro…le N H O L cannot be sustained as a
PBE of this game.
NH 1 OH q
0,0
R
p
-1,0
2
2,-1
(1-p) A
0,0
NL 1 OL (1-q) R
-1,0
Given this strategy from the …rm, the worker’s beliefs after observing
a job o¤er must be
p 1
q= =1
p 1 + (1 p) 0
Given these beliefs, it is optimal for the worker to accept the job o¤er,
since 2 > 0.
For future reference, we shade the branch in which the worker accepts
the …rm’s job o¤er, an action that is independent on the …rm’s type
(something that the worker cannot observe).
If the worker accepts, then the high-quality …rm prefers to make a job
o¤er, OH, since 2 > 0, as prescribed by this strategy pro…le.
The low-quality …rm also prefers to make a job o¤er, O L , which
contradicts the initially proposed strategy pro…le O H N L .
Then, the separating strategy pro…le O H N L cannot be sustained as a
PBE of this game either.
A 2,2
NH 1 OH q
0,0
R
p
-1,0
2
2,-1
(1-p) A
0,0
NL 1 OL (1-q) R
-1,0
Given this strategy from the …rm, the worker’s beliefs after observing
a job o¤er must be
p 1
q= =p
p 1 + (1 p) 1
Intuitively, this implies that after observing O, the worker cannot infer
additional information about the …rm’s quality, since all …rms make
job o¤ers. (Note that this result radically di¤ers from that in the two
separating strategy pro…les described above.)
Given these beliefs, it is optimal for the worker to accept the job o¤er
if EU2 (A) > EU2 (R ), where
EU2 (A) = 2p + (1 p )( 1) = 3p 1
EU2 (R ) = 0 + 0
! 3p 1 = 0
1
p = for A = R
3
1
p > for A > R
3
1
p < for A < R
3
Hence, if p > 13 , the worker accepts the job o¤er and we can shade
branch A in the …gure.
Otherwise, the worker rejects the job o¤er and we must shade branch
R.
If p < 13 , the worker rejects the job o¤er. The high-quality …rm does
not want to make a job o¤er since making it yields a payo¤ of -1 (since
the o¤er is rejected) whereas not making such an o¤er yields a payo¤
of 0.
Hence, the pooling strategy pro…le O H O L cannot be sustained as PBE
of this game when p < 13 .
A 2,2
NH 1 OH q
0,0
R
p
-1,0
2
2,-1
(1-p) A
0,0
NL 1 OL (1-q) R
-1,0
Given this strategy from the …rm, the worker’s beliefs after observing
a job o¤er must be
p 0 0
q= =
p 0 + (1 p) 0 0
Given these beliefs, it is optimal for the worker to accept the job o¤er
if EU2 (A) > EU2 (R ), where
EU2 (R ) = 0
! 2q + ( 1)(1 q) = 0
1
q= for A = R
3
1
q < for R > A
3
1
q > for R < A
3
Hence, if q > 13 , the worker accepts the job o¤er and we can shade
branch A in the …gure.
Otherwise, the worker rejects the job o¤er and we must shade branch
R.
If q < 31 , the worker rejects the job o¤er. The high-quality …rm does
not want to make a job o¤er: making it yield a payo¤ of -1 (since the
o¤er is rejected) whereas not making such an o¤er yields a payo¤ of 0.
Similarly for the low-quality …rm.
Hence, no type of …rm makes job o¤ers and the pooling strategy pro…le
N H N L can be sustained as PBE of this game when q < 13 .
Nature
Low High
Moderate
Applicant 9 4 15 10 11 6 15 10 13 8 15 10
Employer 4 7 4 7 6 7 6 7 14 7 14 7
The …gure on the next slide shades the branches corresponding to the
strategy pro…le we need to check, where the applicant goes to college
both when his innate ability is moderate and high, but does not go to
college when his ability is low.
Nature
Low High
Moderate
Applicant 9 4 15 10 11 6 15 10 13 8 15 10
Employer 4 7 4 7 6 7 6 7 14 7 14 7
1 1
6 1
1 µ2 high intellect with probability: 1 1
=
2 1+ 6 1 4
1 1 1 1
EUF (C ) = 2 7+ 6 7=7
1 1+ 1 1 1 1+ 1 1
2 6 2 6
Clearly, EUF (M) > EUF (C ), implying that if the student did go to
college, the employer hires him as a manager.
The employer’s beliefs are consistent with respect to the student not
going to college, since only when his type is low intellect does he not
go to college.