You are on page 1of 8

Lectures 12-13

Incomplete Information
Static Case
14.12 Game Theory
Muhamet Yildiz

Road Map
1. Examples
2. Bayes rule
3. Definitions
1. Bayesian Game
2. Bayesian Nash Equilibrium

4. Mixed strategies, revisited


5. Economic Applications
1. Cournot Duopoly
2. Auctions
3. Double Auction

Incomplete information
We have incomplete (or asymmetric)
information if one player knows something
(relevant) that some other player does not
know.

An Example
Firm

W ork

H ire
Shirk

H igh p
D o not
hire

N ature

Low 1-p

(1, 2)

H ire

(0, 0)
W

W ork
Shirk

D o not
hire

(0, 1)

(1, 1)

(-1, 2)

(0, 0)

The same example


Work

(1, 2)

Shirk

(0, 1)

W
Nature

High p

Hire

hire

Firm

Work

Low
1-p

(1, 1)

W
Shirk

Do not

(-1, 2)

(0, 0)

Another Example Buy

(p, 2-p)

B
Seller

Dont

High 0.5

Buy

Nature

Low .5

What would you ask


if you were to choose
p from [0,4]?

Dont

Buy

(0, 0)
(p,2-p)

(0, 0)
(p, 1-p)

p
Dont
p

(0,0)

Buy

(p, 1-p)

Dont

(0, 0)

Same Another Example


Buy

(p, 2-p)

B
High 0.5
Dont

Nature

Buy

Low .5

Seller
Nature High 0.5

What would you ask


if you were to choose
p from [0,4]?

Dont

Buy
Dont

Low .5

(0, 0)
(p,2-p)

(0, 0)
(p, 1-p)

(0,0)

Buy

(p, 1-p)

Dont

(0, 0)

Bayes Rule
Prob(A and B)
Prob(A|B) =
Prob(B)
Prob(A and B) = Prob(A|B)Prob(B) = Prob(B|A)Prob(A)

Prob(B|A)Prob(A)
Prob(A|B) =
Prob(B)

Example
Work

p
Success
1-p

Prob(Work|Success) =
p/[p + (1)(1-p)]
Prob(Work|Failure) =
(1-)p/[(1p) + (1)p]

1-p
Shirk
1

Failure

P(W|S)

0.9
0.8

P (w|S ),P (W|F)

0.7
0.6
0.5

P(W|F)

0.4
0.3
0.2
0.1
0

0.2

0.4

0.6

0.8

Bayesian Game (Normal Form)


A Bayesian game is a list
G = {A1,,An;T1,,Tn;p1,,pn;u1,,un}
where
Ai is the action space of i (ai in Ai)
Ti is the type space of i (ti)
pi(t-i|ti) is is belief about the other players
ui(a1,,an;t1,,tn) is is payoff.

An Example
Firm

Work

(1, 2)

Hire
Shirk

High p

(0, 1)

Do not (0, 0)
hire

Nature

Low 1-p

Hire

Work
Shirk

Do not
hire

TFirm={tf};
TW = {High,Low}
AFirm = {Hire, Dont}
AW = {Work,Shirk}
pF(High) = p
pF(Low) = 1-p

(1, 1)

(-1, 2)

(0, 0)

Bayesian Nash equilibrium


A Bayesian Nash equilibrium is a Nash equilibrium of
a Bayesian game.
Given any Bayesian game G =
{A1,,An;T1,,Tn;p1,,pn;u1,,un}
a strategy of a player i in a is any function si:Ti Ai;
A strategy profile s* = (s1*,, s1*) is a Bayesian Nash
equilibrium iff si*(ti) solves

max u (s (t ),..., s (t ), a , s (t ),..., s (t ); t )p (t


ai Ai

t i T i

*
1

*
i 1

i 1

*
i +1

i +1

*
n

| ti )

i.e., si* is a best response to s-i*.

An Example
Firm

Work

(1, 2)

Hire
Shirk

High p

(0, 1)

Do not (0, 0)
hire

Nature

Low 1-p

Hire

Work
Shirk

Do not
hire

(0, 0)

TFirm={tf};
TW = {High,Low}
AFirm = {Hire, Dont}
AW = {Work,Shirk}
pF(High) = p >1/2
pF(Low) = 1-p

(1, 1)

(-1, 2)

sF* = Hire,
sF* (High) = Work
sF* (Low) = Shirk
Another equilibrium?

Stag Hunt, Mixed Strategy

(2,2)

(4,0)

(0,4)

(6,6)

Mixed Strategies
t and v are iid with uniform
distribution on [,].
t and v are privately known by
1 and 2, respectively, i.e., are
types of 1 and 2, respectively.
Pure strategy:
2+t,2+v 4+t,0
s1(t) = Rabbit iff t > 0;
s2(v) = Rabbit iff t > 0.
0,4+v
6,6
p = Prob(s1(t)=Rabbit|v) =
Prob(t > 0) = 1/2.
U1(R|t) = t +2q+4(1-q) = t + 4 2q q = Prob(s2(v)=Rabbit|t) =
U1(S|t) = 6(1-q);
1/2.
U1(R|t) > U1(S|t) t+42q > 6(1-q)
t > 6-6q+2q-4 = 2 4q = 0.