Applications of dynamic games: Bargaining and reputation · 28.10.2014 Daniel Spiro, ECON3200/4200,...
Transcript of Applications of dynamic games: Bargaining and reputation · 28.10.2014 Daniel Spiro, ECON3200/4200,...
28.10.2014 Daniel Spiro, ECON3200/4200, Lecture 4 1
Applications of dynamic games:
Bargaining and reputation
Lectures in Game Theory
Fall 2014, Lecture 4
28.10.2014 Daniel Spiro, ECON3200/4200, Lecture 4 2
Modeling of
negotiations between parties with opposing
interests:
Sequential bargaining games
non-cooperative cooperation between parties
that in the short run want to deviate:
Repeated games
28.10.2014 Daniel Spiro, ECON3200/4200, Lecture 4 3
d: default outcome,
disagreement point,
2uBargaining
d
1u
Agreement
creates value, but
how to divide?
A
B
A, B: outcomes
A: efficient outcome if
transfers are possible
Efficient bargaining:
Agree on efficient out-
come and divide surplus.
Applications: Bilateral monopoly
Labor market negotiations
International negotiations
28.10.2014 Daniel Spiro, ECON3200/4200, Lecture 4 4
2u
1
The standard
bargaining solution
1u
1
What do the bargaining weights depend on?
First: Normalize.
Then: Divide surplus
according to
bargaining weights, p1
and p2. 1
p
2p
28.10.2014 Daniel Spiro, ECON3200/4200, Lecture 4 5
Ultimatum
bargaining
1
2 A
R
1m
,1
m1
1 m
m1 is 1’s
share. 1
u
2u
1
0 ,0
1
Bargaining weight = 1
for the proposer
28.10.2014 Daniel Spiro, ECON3200/4200, Lecture 4 6
Two-period
alternating offer
1
2 A
R
1m
,1
m1
1 m
2
1 A
R
2m
,2
m )1(2
m
0 ,0
mt is 1’s
share in
period t.
is the dis-
count factor
1u
2u
1
28.10.2014 Daniel Spiro, ECON3200/4200, Lecture 4 7
Three-period
alternating offer
1
2 A
R
1m
,1
m1
1 m
2
1 A
R
2m
,2
m )1(2
m
mt is 1’s
share in
period t.
is the dis-
count factor
1u
2u
2 1
1
28.10.2014 Daniel Spiro, ECON3200/4200, Lecture 4 8
1
2 A
R
1m
,1
m1
1 m
2
1 A
R
2m
,2
m )1(2
m
1
Backward induction:
Period 3: 1 suggests m3 =1, 2 will
accept. Payoff is , 0 2
Period 2: 1 indiff between accepting
and rejecting if m2 = . Hence 2 will
suggest m2=, 1 will accept. Payoff is
and (1-m2) = (1- )
2
Period 1: 2 indiff between accepting and rejecting if 1-m1 = (1- ) . Hence, if 1 will suggest m1=1(1) then 2
will accept. Will 1 want to suggest this? Yes since m1 > 2
2
SPNE: s1=[m1=1(1), A if m2, m3 =1]
s2=[A if m1 1(1), m2=, A if m3 0]
28.10.2014 Daniel Spiro, ECON3200/4200, Lecture 4 9
Three-period
alternating offer
1
2 A
R
1m
,1
m1
1 m
2
1 A
R
2m
,2
m )1(2
m
mt is 1’s
share in
period t.
is the dis-
count factor
1u
2u
2 1
1
28.10.2014 Daniel Spiro, ECON3200/4200, Lecture 4 10
Rubinstein’s
bargaining model
1
2 A
R
1m
,1
m1
1 m
2
1 A
R
2m
,2
m )1(2
m
,3
2m )1(3
2 m
mt is 1’s
share in
period t.
is the dis-
count factor
1u
2u
2 13
2m
Equal bargaining
weight for pl. 1
in periods 1 & 3.
Finding the solution of Rubinstein’s bar-
gaining model with an infinite time horizon
)1(121
mm =
Equal bargaining weight for P1 in periods 1 and 3:
31mm =
Three equations in three unknown. Solution:
In period 1, P1 makes the best proposal that P2
accepts:
3
2
2mm =
In period 2, P2 makes the best proposal that P1
accepts:
==
1
131
mm
=
12
m
28.10.2014 Daniel Spiro, ECON3200/4200, Lecture 4 12
Solution of Rubinstein’s bargaining model
with an infinite time horizon
1tm only if and ifaccepts 1
Unique subgame perfect Nash equilibrium:
otherformandherselfform tt
=
=1
111
1
1 suggests 1
=
1
11
1)-(1
tm only if and ifaccepts 2
=
=
=
1
11
1
11
1tt mm suggests 2
1st mover advantage: 1’s share is larger than 2’s.
1’s proposal at time t = 1 is accepted.
If disc. factors are different: It pays to be patient.
28.10.2014 Daniel Spiro, ECON3200/4200, Lecture 4 13
Repeated games with an infinite time horizon
D
D
C
C
1 1,
2 2,3 0,
0 3,
Prisoners’
dilemma
2u
1u
1 2 3
1
2
3
Can repetition discipline
the players to cooperate?
Deviating yields a short run gain.
Deviating yields a loss of reputation
that undermines future cooperation.
Yes, if gain
now PV of
future loss
28.10.2014 Daniel Spiro, ECON3200/4200, Lecture 4 14
Cooperation in infinitely rep.
.) with(Start .play otherwise
earlier; used been has ifPlay
:strategy"Trigger "
CC
DD
If cooperation breaks down, it will never be restarted.
Prisoners’ Dil.
Subgame perfect Nash equilibrium (NE in all subgames)?
If cooperation has broken down: NE in the subgames.
If cooperation has not broken down:
Short-run gain PV of long-run loss
)12(1
23
21
2
1
2u
1u
1 2 3
1
2
3
28.10.2014 Daniel Spiro, ECON3200/4200, Lecture 4 15
“Getting Even”
period. previous inperm"" w/o played
opponent the ifplay to Permitted""
.play to permitted"" unless ,Play
D
D
DC
If 1 deviates, then 1 is punished in the next period.
If cooperation has not broken down:
Short-run gain PV of loss in next period
)02(23 21 2
1
If cooperation has broken down:
)02(01 21 2
1
2u
1u
1 2 3
1
2
3
28.10.2014 Daniel Spiro, ECON3200/4200, Lecture 4 16
Repeated games– Economic lessons
Two sources of instability: 1) temptation for deviation in coop stage and
2) temptation to deviate when the player is being punished. With the
numbers in previous example the two factors are equally strong.
Other numbers may change the relative force.
”Getting even” and ”trigger strategy” are ways of upholding
cooperation. In previous example they happen to be equally
stable. Usually they are not.
28.10.2014 Daniel Spiro, ECON3200/4200, Lecture 4 17
Repeated games– Economic lessons
The results hinge on the game being infinitelly repeated. Finite PD:
• In last period, DD is only NE no matter of history.
• Therefore payoffs by actions in second to last period is indep of
payoff in last period.
• Hence, in second to last period, DD is only NE
• Similar backward induction leads to DD in all periods.
Further applications
28.10.2014 Daniel Spiro, ECON3200/4200, Lecture 4 18
In examples, the will to punish the others was a non-issue. In
many economic settings there is a cost, need to check that
punishment is a NE.
Cooperation for climate control Based on Barrett, Journal of Theoretical Politics 11 (1999), 519–541.
Cooperate: Reduce the emis-
sion of greenhouse gases.
Defect: Do not reduce the
emission of greenh. gases.
( )C n c dn =
In total N countries, where
n countries cooperate.
bnnD
= )(
Payoff function:
Conditions: 0 bdc
( 1) 0b N c dN
Cooperation for climate control (cont.)
An agreement between k participating countries based on:
“Getting Even”: A participating country plays Cooperate
by reducing emission, except if another participating
country has been the sole deviator from “Getting Even”
in the previous period (in which case, the play of Defect in
the sense of no reduction of emission is specified).
Non-participating countries play Defect after any history.
Two observations:
– Not a subgame perfect NE if k is too low.
– Not “renegotiation-proof ” if k is too high.
Subgame perfect Nash equilibrium (NE in all subgames)?
Cooperation for climate control (cont.)
If cooperation has not broken down:
Short-run gain PV of loss in next period
( ) )()()1( cdkcdcdkkb dbd
dck
1
If cooperation has broken down:
( ) )()(0 cdkcdcd d
dck
1
Are the punishing countries better off during punishment (so
they do not want to return to the original equilibrium at once)?
cdkb d
cbk
:1 to close & db =
d
ck
d
c 1