Scribd
Upload
63 views24 pages

Midterm Review 2: Introduction To Game Theory

The midterm exam will cover backward induction, SPNE, ICD, marginal contribution, and the core. It will take place on December 7th from 11:30-12:30 in Room 310. Students should bring their own calculator. Topics covered include game trees, subgame perfect Nash equilibrium, repeated games, signaling, and cooperative theory.

Uploaded by

Muhammad Amin
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
63 views24 pages

Midterm Review 2: Introduction To Game Theory

The midterm exam will cover backward induction, SPNE, ICD, marginal contribution, and the core. It will take place on December 7th from 11:30-12:30 in Room 310. Students should bring their own calculator. Topics covered include game trees, subgame perfect Nash equilibrium, repeated games, signaling, and cooperative theory.

Uploaded by

Muhammad Amin
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/ 24

Midterm Review 2

Introduction to Game Theory

New York University Shanghai

Fall 2017
Midterm 2

I Dec 7th (Th), 11:30-12:30

I Room 310.

I 4 problems, covering Backward Induction, SPNE, ICD, Marginal


Contribution, Core.
(Applications: Bargaining, Signaling, Forward induction, Repeated
games)

I Bring your own calculator.


What we have covered

I Game tree, information set

I Backward Induction, Bargaining

I Subgame Perfect Nash equilibrium, Repeated games

I Iterated Conditional Dominance, Signaling, Forward Induction

I Cooperative Theory
Information set

I A player only knows which information set has been reached, but
cannot differentiate between nodes within the information set.

I An information set can contain only one node.

I If all the information sets of a game are singletons, then this is a


game of perfect information.
Information set

h0 : US active. h1 : Russia active. h2 : US active. h20 : Russia active.


Backward Induction

I BI can only be applied in finite games with perfect information.

I Application: Bargaining
Bargaining

I Ann and Bob bargain over a pie of 100. (xa , xb )


I Discount factors: δa , δb
I 1st round A offers. 2nd round B offers. 3rd round A offers.

I 3rd round: B accepts any xb ≥ 0. So A offers (100, 0).

I 2nd round: A accepts any xa ≥ 100δa .


So B offers (100δa , 100(1 − δa )).

I 1st round: B accepts any xb ≥ 100(1 − δa )δb .


So A offers (100(1 − δa )(1 − δb ), 100(1 − δa )δb ).
Bargaining

I Ann and Bob bargain over a pie of 100. (xa , xb )


I Discount factors: δa , δb
I 1st round A offers. 2nd round B offers. 3rd round A offers.

I 3rd round: B accepts any xb ≥ 0. So A offers (100, 0).

I 2nd round: A accepts any xa ≥ 100δa .


So B offers (100δa , 100(1 − δa )).

I 1st round: B accepts any xb ≥ 100(1 − δa )δb .


So A offers (100(1 − δa )(1 − δb ), 100(1 − δa )δb ).
Bargaining

I Ann and Bob bargain over a pie of 100. (xa , xb )


I Discount factors: δa , δb
I 1st round A offers. 2nd round B offers. 3rd round A offers.

I 3rd round: B accepts any xb ≥ 0. So A offers (100, 0).

I 2nd round: A accepts any xa ≥ 100δa .


So B offers (100δa , 100(1 − δa )).

I 1st round: B accepts any xb ≥ 100(1 − δa )δb .


So A offers (100(1 − δa )(1 − δb ), 100(1 − δa )δb ).
Bargaining

I Ann and Bob bargain over a pie of 100. (xa , xb )


I Discount factors: δa , δb
I 1st round A offers. 2nd round B offers. 3rd round A offers.

I 3rd round: B accepts any xb ≥ 0. So A offers (100, 0).

I 2nd round: A accepts any xa ≥ 100δa .


So B offers (100δa , 100(1 − δa )).

I 1st round: B accepts any xb ≥ 100(1 − δa )δb .


So A offers (100(1 − δa )(1 − δb ), 100(1 − δa )δb ).
Subgame

Definition
I A subgame starts with a single node and contains all its successors.
I If a node is in a subgame, so do all the other nodes from the
information set containing that node.

Note
I The boundary of a subgame should not cut through an information
set.
I The original game is a subgame of itself.
I This game has one subgame.
Subgame Perfect Nash Equilibrium

I A strategy profile (sa∗ , sb∗ ) is subgame perfect if it represents a Nash


equilibrium in every subgame of the original game.

I Single Deviation Principle: (sa∗ , sb∗ ) is subgame perfect if there exists


no profitable one-shot deviation.

I Identifying all pure strategy SPNE in a finite game:


I Step 1: Use matrix representation to find all Nash equilibria.
I Step 2: Use Single Deviation Principle to find all SPNE.
I This is a SPNE.
I Given that the other player doesn’t deviate, there is no profitable
deviation in any single stage.
I Not a SPNE.
I Given that the other player doesn’t deviate, Russia has a profitable
deviation in the second stage.
SPNE in Infinitely Repeated Games

I Identify all types of histories in which players are specified different


actions
I Eg 1. For Grim Trigger, consider the two cases:
I The two players cooperated in all previous periods (c,c,c,c...)
I otherwise (p, p, p, p...)
I Eg 2. Consider the four cases in the limited punishment question:
I Last period was (H, H), or last three periods was punishment
(c, c, c, c....)
I first period of punishment (p, p, p, c....)
I second period of punishment (p, p, c, c....)
I third period of punishment (p, c, c, c....)
I For each case and each player, check what is the most profitable
deviation and whether the player wants to deviate.
Iterated Conditional Dominance

I Step 1: In the original game G 0 , for every information set h, identify


all the dominated strategies in G 0 (h) for the player who is active at
h.
I Step 2: Eliminate all the strategies identified in Step 1 in all
information sets.
This gives you the reduced game G 1 .

I Repeat Steps 1 and 2 in G 1 , which will give you the reduced game
G 2.
I Repeat the process until no more strategies can be eliminated in
this way.

I Note: in each round of elimination, always finish Step 1 first, and


then perform Step 2.
Eliminate A for P1 and D for P2.
However, if we first eliminate A in both information sets, then D cannot
be identified as a dominated strategy.
Signaling: mimicing High type is a dominated strategy for Low type
Forward Induction: my opponent must be rational in previous rounds

I The only rational explanation for Ann giving up 1.5 is that she
wants to get 2 in the subgame G2.
I So if she has chosen Movie, she definitely would choose K in G2.
Bob would not choose T and Ann should not worry about him
choosing T. (DT, T) is not a good prediction.
Cooperative Theory

I Marginal Contribution Principle implies

xi ≤ v (N) − v (N \ {i}) (1)

e.g. x1 ≤ v ({1, 2, 3, 4}) − v ({2, 3, 4})


I Core implies
Σi∈S xi ≥ v ({S}) (2)
e.g. x1 + x2 ≥ v ({1, 2})
I Core also implies

Σi∈S xi ≤ v (N) − v (N \ S) (3)

e.g. x3 + x4 ≤ v ({1, 2, 3, 4}) − v ({1, 2})


I (3) is an extension of (1). (3) is equivalent to (2).
Cooperative Theory

I Marginal Contribution Principle implies

xi ≤ v (N) − v (N \ {i}) (1)

e.g. x1 ≤ v ({1, 2, 3, 4}) − v ({2, 3, 4})


I Core implies
Σi∈S xi ≥ v ({S}) (2)
e.g. x1 + x2 ≥ v ({1, 2})
I Core also implies

Σi∈S xi ≤ v (N) − v (N \ S) (3)

e.g. x3 + x4 ≤ v ({1, 2, 3, 4}) − v ({1, 2})


I (3) is an extension of (1). (3) is equivalent to (2).
Cooperative Theory

I Marginal Contribution Principle implies

xi ≤ v (N) − v (N \ {i}) (1)

e.g. x1 ≤ v ({1, 2, 3, 4}) − v ({2, 3, 4})


I Core implies
Σi∈S xi ≥ v ({S}) (2)
e.g. x1 + x2 ≥ v ({1, 2})
I Core also implies

Σi∈S xi ≤ v (N) − v (N \ S) (3)

e.g. x3 + x4 ≤ v ({1, 2, 3, 4}) − v ({1, 2})


I (3) is an extension of (1). (3) is equivalent to (2).

You might also like