0% found this document useful (0 votes)
107 views34 pages

Week 6 - Chapter 29 - Game Theory PDF

The document discusses game theory and the prisoner's dilemma. It provides an example of the prisoner's dilemma involving two suspects, Raj and Howard. It presents the game in normal form and analyzes the Nash equilibrium, where both players confess even though they would be collectively better off if neither confessed. Repeated games are introduced, where the same game is played multiple times, allowing players to retaliate or reward each other's past actions. Finitely repeated games are discussed, where players seek to maximize total payoff over a set number of periods. The Nash reversion strategy is presented, where players cooperate but cheat permanently if either player deviates.

Uploaded by

Nishant Naidu
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
107 views34 pages

Week 6 - Chapter 29 - Game Theory PDF

The document discusses game theory and the prisoner's dilemma. It provides an example of the prisoner's dilemma involving two suspects, Raj and Howard. It presents the game in normal form and analyzes the Nash equilibrium, where both players confess even though they would be collectively better off if neither confessed. Repeated games are introduced, where the same game is played multiple times, allowing players to retaliate or reward each other's past actions. Finitely repeated games are discussed, where players seek to maximize total payoff over a set number of periods. The Nash reversion strategy is presented, where players cooperate but cheat permanently if either player deviates.

Uploaded by

Nishant Naidu
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 34

GAME

THEORY

Week 6
(Chapter 29; except 29.3)

1
Part 1
SIMULTANEOUS-MOVE GAMES

2
Prisoner’s Dilemma

•  Two suspects of a crime, Raj and Howard are


arrested and are held in separate cells

•  Each of them is privately told that


–  If he confesses and the other person does not, then he will
be rewarded with a light sentence of 1 year in jail and the
other person will go to jail for 10 years.
–  If both confess, each will get 5 years in jail
–  If neither confesses, each will get 2 years in jail

3
Normal Form Representation of the Game

•  Players: Raj and Howard


•  Strategies: “Confess” and “Do Not Confess”
•  Outcomes and Payoffs:
–  If only one person confesses, then he will be rewarded
with a light sentence of 1 year in jail and the other person
will go to jail for 10 years.
–  If both confess, each will get 5 years in jail
–  If neither confesses, each will get 2 years in jail

4
Normal Form Representation of the Game

Howard
Confess Do Not Confess
Confess -5, -5 -1, -10
Raj
Do Not Confess -10, -1 -2, -2

Raj’s payoff if he does not confess Howard’s payoff if he does


but Howard does not confess but Raj does

5
Assumptions
•  Every player is rational
–  Everyone knows that everyone is rational
–  Everyone knows that everyone knows that everyone is
rational…
–  (Rationality is common knowledge)

•  Everyone knows the structure of the game
–  Everyone knows that everyone knows the structure…
–  (Structure of the game is common knowledge)

6
What is the likely outcome of the game?

•  What is the “best” choice for each player?

•  Raj’s best response is the strategy that maximizes


Raj’s payoff given the strategy chosen by Howard
–  Given any strategy chosen by Howard, Raj has a best
response
–  [Given any strategy chosen by Raj, Howard has a best
response]

7
Best Response in Prisoner’s Dilemma

Howard
Confess Do Not Confess
Confess -5, -5 -1, -10
Raj
Do Not Confess -10, -1 -2, -2

Raj’s best response if


Howard plays Confess Howard’s best response if Raj
plays Do Not Confess

8
Nash Equilibrium

•  Each player chooses a strategy that maximizes his


payoff given the strategy chosen by the other player
–  Each player selects his best response to the strategy
actually chosen by the other player

•  The Nash equilibrium strategies are the strategies


that represent the players’ mutual best responses

9
John Nash (1928-2015)

10
Nash Equilibrium in Prisoner’s Dilemma

Howard
Confess Do Not Confess
Confess -5, -5 -1, -10
Raj
Do Not Confess -10, -1 -2, -2

11
Interpreting Nash Equilibrium

•  At Nash equilibrium, no one has an incentive to


deviate to another strategy
–  When Howard chooses “Confess”, Raj has no incentive to
choose “Do Not Confess”

•  “No regret” at Nash equilibrium


–  Looking back at his decision, Raj will not regret
–  Given that Howard chooses “Confess”, Raj should indeed
choose “Confess”

12
Implications of Prisoner’s Dilemma

•  Raj and Howard will be collectively better off if they do not


confess
–  If they do not confess, each will get 2 years in jail

•  But (Do not confess, Do not confess) is not a Nash


equilibrium!

•  Players’ rational pursuit of their individual best interest can


lead to outcomes that are bad for everyone

•  Individual rationality does not bring about social optimum (in


this case)
13
Implications of Prisoner’s Dilemma

•  But First Welfare Theorem suggest that individual rationality


brings about social optimum

•  It is not from the benevolence of the butcher, the brewer, or
the baker that we expect our dinner, but from their regard to
their own interest.
–  Adam Smith

•  What drives the different conclusions?


First Welfare Theorem assumes that there is no strategic
interactions (all players are price takers).

14
Prisoner’s Dilemma More Generally
Player 2
Cheat Cooperate
Cheat C, C A, D
Player 1
Cooperate D, A B, B

Where A>B>C>D

For a game to be a Prisoner’s dilemma,
•  “Cheat” is a strictly dominant strategy for all players
•  But it is socially optimal to cooperate
15
Bertrand Competition as
a Prisoner’s Dilemma

•  Two firms with same MC set price once and for all

•  The only Nash equilibrium is to set P=MC

•  Each firm can get much higher profit by charging the


monopoly price
–  But this is not a Nash equilibrium

16
Part 2
REPEATED GAMES

17
Intuition of Repeated Games

•  In a one-shot game, players do not need to worry about


the “consequence” of their actions
–  Retaliation is impossible
–  No punishment for cheating

•  If players interact repeatedly, then one could base their


action on what the other player has done to him in the
previous periods

•  It may be possible that no one will cheat

•  Repeated games may repeat finitely or infinitely


18
Finitely Repeated Game

•  The same players repeat the same static games for T periods

•  The static game in each period is called the stage game

•  The players observe the outcomes of all previous stage games

•  Every player maximizes the sum of the discounted payoff in


all periods

19
Finitely Repeated Prisoner’s Dilemma
Robin
Cheat Cooperate
Cheat 3, 3 10, 0
Wayne
Cooperate 0, 10 6, 6

•  Suppose each player maximizes his total payoff

•  The game is repeated 100 times

20
Nash Reversion Strategy

•  Wayne’s strategy
–  I will start off choosing “cooperate”
–  If in the last period, either player choses “cheat”, then
from this period on, I will choose “cheat” in every period

•  Robin uses the same strategy

•  One period of cheating will trigger permanent


punishment – reverting to the Nash equilibrium
strategy in the one-shot game
21
Should Wayne cheat in period 1?

•  Suppose the discount factor of Wayne is δ=1


•  If Wayne cooperates in every period left, his total payoff is
6*100 = 600
•  If Wayne cheats in period 1
–  He gets 10 in that period
–  But from period 2 on, both players will cheat
–  Wayne gets 3 from period 2 on
–  Wayne’s total payoff if cheating in period 1 is
10 + 3*99 = 307 <600
•  So Wayne should cooperate?
–  This line of reasoning is flawed!
22
Should Wayne cheat in period 100?

Robin
Cheat Cooperate
Cheat 3, 3 10, 0
Wayne
Cooperate 0, 10 6, 6

•  This is the last stage game


•  Game over after period 100
•  It is as if they are playing the static game
•  Both players will cheat in period 100

23
Should Wayne cheat in period 99?

•  By backward induction
•  Every player knows that each player will cheat in period
100
•  Hence, no gain by choosing to cooperate in period 99.
Every player chooses “cheat”
•  By the same reasoning, every player chooses “cheat” in
period 98
•  …
•  Every player chooses “cheat” in period 1

24
Why cooperation cannot be sustained?

•  In this case, Nash reversion strategy is not credible


–  In the last period every player will cheat, regardless of
what have been chosen in previous period
–  The “end game” ruins cooperation

•  If the stage game is a prisoner’s dilemma,


cooperation cannot be sustained if game is only
finitely repeated

25
Infinitely Repeated Games

•  Suppose players play the same static game for


infinitely many times

•  Each player maximizes the sum of discounted payoff


over all periods

•  Assume each player’s discount factor is δ=0.9

26
No Cheating if players are patient

•  Assume each player’s discount factor is δ=0.9

•  If Wayne deviates in period y


–  He gets 10 in that period
–  But 3 in every period after
–  Total payoff is 10+0.9(3)+0.92(3)+0.93(3)+…=37

•  If he cooperates, he gets
6+0.9(6)+0.92(6)+0.93(6)+…=60

•  Wayne should cooperate


27
Cheating if players are impatient

•  Assume each player’s discount factor is δ=0.1

•  If Wayne deviates in period y


–  He gets 10 in that period
–  But 3 in every period after
–  Total payoff is 10+0.1(3)+0.12(3)+0.13(3)+…=10.33

•  If he cooperates, he gets
6+0.1(6)+0.12(6)+0.13(6)+…=6.67

•  Wayne should cheat
28
Part 3
SEQUENTIAL GAMES

29
Predation Game

•  A market currently has one incumbent firm

•  An entrant chooses “out” or “in”

•  The incumbent observes the entrant’s choice

•  If the entrant chooses “in”, the incumbent can choose to


“fight” or “accommodate”

•  If the entrant chooses “out”, the incumbent does nothing


–  Incumbent’s strategy is a function of entrant’s strategy

30
Extensive Form Representation of the
Predation Game
Entrant
Out In
Incumbent
! Payoff $ ! $
# Entrant &=# 0
# Payoff # &&
& " 2 % Fight Accommodate
" Incumbent %

" −3 % ! 2 $
$ ' # &
# −1 & " 1 %

31
Backward Induction

•  Solve for the optimal strategy in the last stage

•  Solve for the optimal strategy in the second last


stage

•  Solve for the optimal strategy in the third last stage…

32
Solving the Predation Game
By Backward Induction
Entrant
Out In
Incumbent

! 0 $
# & Fight Accommodate
" 2 %

" −3 % ! 2 $
$ ' # &
# −1 & " 1 %
If Entrant enters, Incumbent’s optimal strategy is
to accommodate
33
Solving the Predation Game
By Backward Induction
Entrant
Out In
Incumbent

! 0 $
# & Fight Accommodate
" 2 %

" −3 % ! 2 $
$ ' # &
# −1 & " 1 %
Entrant should enter, knowing that once it enters,
incumbent’s optimal strategy is to accommodate
34

You might also like