lOMoARcPSD|32983522

Sample/practice exam 2019, questions and answers

Advanced Game Theory (Royal Holloway, University of London)

Studocu is not sponsored or endorsed by any college or university

Downloaded by AKSHITI KACHHAWAH (2020.akshiti.kachhawah@ves.ac.in)
 lOMoARcPSD|32983522

Exam Practice Questions

Alexander Vickery
EC3324 - Topics in Game Theory
Due: May 13, 2019

Answer the questions on a separate sheet. Your work can be scanned and sent to me, or if
you prefer you can print it and bring to my office hours. This worksheet has 10 question,
for a total of 175 points.

Name:

Question Marks Score

Simultaneous Move Incomplete Information Game 25

Location Game 15

Signaling Game 25

Incomplete Information Game 20

Pure and Mixed Strategies 15

Three Player Normal Form Game 15

Perfect Bayesian Equilibrium (PBE) 15

Cournot Duopoly 20

Announcing Integers 10

Extensive Form Game 15

Total: 175

Downloaded by AKSHITI KACHHAWAH (2020.akshiti.kachhawah@ves.ac.in)

 lOMoARcPSD|32983522

EC3324 - Topics in Game Theory Exam Practice Questions - Spring 2019

Problem 1: Simultaneous Move Incomplete Information Game

Two record labels are required to make simultaneous decisions regarding the release of their upcoming
albums. Their payoffs are given by the following matrix:

YG Entertainment

D(elay) R(elease)

D(elay) −c1 , −c2 1 − c1 , −2

Cube Entertainment
R(elease) −2, 1 − c2 0, 0

where D denotes delaying and R denotes releasing the album. ci denotes the cost of delaying for record
label i. This cost is label i’s private information. Assume that ci takes two possible values, 0.25 and 1.25.
The probability that ci = 1.25 is p and therefore the probability that ci = 0.25 is (1 − p). In addition, the
realization of ci is independent across record labels.

Please answer the following questions.

(a) 10 Marks How many “types” does each record label have? How many strategies does each record
label have? Substitute in the possible values for ci and represent this interaction as an incomplete
information game.

Solution:
Each record label has two “types” - given by their (independent) draws of ci . This means that each
country’s strategy defines an action for each “type”. Specifically the strategy set Si for each player
i is:
Si = {(D, D), (D, R), (R, D), (R, R)} for i = 1, 2.
So each label has 4 strategies.

The incomplete information game can be represented by:

YG: c2 = 0.25 YG: c2 = 1.25
D R D R
Cube: D −0.25, −0.25 0.75, −2 Cube: D −0.25, −1.25 0.75, −2
c1 = 0.25 R −2, 0.75 0, 0 c1 = 0.25 R −2, −0.25 0, 0

YG: c2 = 0.25 YG: c2 = 1.25

D R D R
Cube: D −1.25, −0.25 −0.25, 2 Cube: D −1.25, −1.25 −0.25, −2
c1 = 1.25 R −2, −0.75 0, 0 c1 = 1.25 R −2, −0.25 0, 0

(b) 5 Marks If p = 1/2, is there a Bayesian Nash equilibrium in which each label always chooses D,
regardless of its cost ci ?

Solution:
Since the game is completely symmetric we only need to check that the strategy profile (D, D) is
a best response to the other label choosing (D, D) for one label. The expected payoff from the

Downloaded by AKSHITI KACHHAWAH (2020.akshiti.kachhawah@ves.ac.in)

 lOMoARcPSD|32983522

EC3324 - Topics in Game Theory Exam Practice Questions - Spring 2019

strategy profile (D, D) for label i when label −i is choosing (D, D) is:

Ei [(D, D)|(D, D)] = −3/4

Therefore, for the strategy profile (D, D) to form part of a BNE, it must be true that:

Ei [(D, D)|(D, D)] ≥ Ei [(D, R)|(D, D)] = −9/8

Ei [(D, D)|(D, D)] ≥ Ei [(R, D)|(D, D)] = −13/8

and
Ei [(D, D)|(D, D)] ≥ Ei [(R, R)|(D, D)] = −8/4
Clearly each of the above conditions are satisfied, therefore if p = 1/2 there is a BNE in which each
label always chooses D regardless of its cost ci , this is given by:

{(D, D), (D, D)}

with p = 1/2.

(c) 10 Marks If p is arbitrary, is there a Bayesian Nash equilibrium in which each record label always
chooses D, regardless of its cost ci ?

Solution:
Since the game is completely symmetric we only need to check that the strategy profile (D, D) is
a best response to the other label choosing (D, D) for one country. The expected payoff from the
strategy profile (D, D) for country i when country −i is choosing (D, D) is:

Ei [(D, D)|(D, D)] = −0.25(1 − p)2 + −0.25p(1 − p) + −1.25(1 − p)p + −1.25p2 = −0.25 − p

Therefore, for the strategy profile (D, D) to form part of a BNE, it must be true that:

Ei [(D, D)|(D, D)] ≥ Ei [(D, R)|(D, D)]

Ei [(D, D)|(D, D)] ≥ Ei [(R, D)|(D, D)]

and
Ei [(D, D)|(D, D)] ≥ Ei [(R, R)|(D, D)]
Calculating the above gives:

Ei [(D, R)|(D, D)] = −0.25(1 − p)2 + −0.25p(1 − p) + −2(1 − p)p + −2p2 = −0.25 − 1.75p

Ei [(R, D)|(D, D)] = −2(1 − p)2 + −2p(1 − p) + −1.25(1 − p)p + −1.25p2 = 0.25(3p − 8)
Ei [(R, R)|(D, D)] = −2(1 − p)2 + −2p(1 − p) + −2(1 − p)p + −2p2 = −2

Clearly each of the above inequalities are satisfied for all values of p, therefore if p is arbitrary there
is still a BNE in which each country always chooses D regardless of its cost ci , this is given by:

{(D, D), (D, D)}

this is true for all p ∈ [0, 1].

Downloaded by AKSHITI KACHHAWAH (2020.akshiti.kachhawah@ves.ac.in)

 lOMoARcPSD|32983522

EC3324 - Topics in Game Theory Exam Practice Questions - Spring 2019

Total Marks for Question 1: 25

Downloaded by AKSHITI KACHHAWAH (2020.akshiti.kachhawah@ves.ac.in)

 lOMoARcPSD|32983522

EC3324 - Topics in Game Theory Exam Practice Questions - Spring 2019

Problem 2: Location Game

Consider the location game, but with a few twists - there are endogenous prices and exogenous locations.
Now (to make things easier), let the location space be the unit interval, [0, 1]. Let the locations of the two
vendors be fixed at the end points of the city: vendor 1 is located at y1 = 0 and firm 2 is located at y2 = 1.
Denote the prices chosen by the two vendors by p1 and p2 , respectively. Assume that the vendors have zero
production costs and that they compete by simultaneously choosing prices. Let the consumers be distributed
uniformly on the unit interval, and let the transportation costs be quadratic, so that the transportation cost
of a consumer located at x when buying from firm i is k(x − yi )2 , for a given constant k > 0. Unlike last
week, however, allow for the possibility that the two firms products also differ in quality: firm i’s product if it
were located at consumer x’s ideal point would be worth Vi to consumer i, where we allow for the possibility
that V1 6= V2 . To summarize, if the consumer located at x buys from firm 1, her utility is:
V1 − p1 − kx2
and if she buys from firm 2, her utility is:
V2 − p2 − k(1 − x)2
Please answer the following questions:

(a) 5 Marks Given p1 and p2 , compute the location of the consumer who is just indifferent between buying
from each of the two vendors - call this location x?

Solution:
The location of the indifferent consumer, denoted x, is given by the solution to:

V1 − p1 − kx2 = V2 − p2 − k(1 − x)2

or
2
k (1 − x) − kx2 = V2 − V1 + p1 − p2 (rearrange)
k − 2kx = V2 − V1 + p1 − p2 (simplify l.h.s)
2kx − k = V1 − V2 − p1 + p2 (multiplying by -1)

or, finally solving,

1 V 1 − V 2 − p1 + p 2
x= + .
2 2k
This expression is the demand for firm 1, while 1 − x is the demand for firm 2. If firms 1 and 2 offer
packages that have the same quality and price, then each firm captures half of the market. Firm
i can capture more than half of the market by either offering a lower price or by offering a higher
quality.

(b) 3 Marks Given your answer in (a), write the demand function and the profit maximization problem
of each firm.

Solution:
Define the demand function for each firm
1 V 1 − V 2 − p1 + p 2
D1 (p1 , p2 ; V1 , V2 ) = x = +
2 2k

Downloaded by AKSHITI KACHHAWAH (2020.akshiti.kachhawah@ves.ac.in)

 lOMoARcPSD|32983522

EC3324 - Topics in Game Theory Exam Practice Questions - Spring 2019

and
1 V 2 − V 1 − p2 + p 1
D2 (p1 , p2 ; V1 , V2 ) = 1 − x = +
2 2k
Since production costs are zero, the profits for firm i is simply its revenue pi Di . Hence, in a
Bertrand-Nash price-setting equilibrium, firm 1 chooses p1 , given p2 , so as to maximize:
 
1 V 1 − V 2 − p 1 + p2
p1 D 1 = p 1 +
2 2k

Firm 2 faces a similar problem in that firm 2 wishes to maximize:

 
1 V 2 − V 1 − p 2 + p1
p2 D 2 = p 2 +
2 2k

(c) 6 Marks Solve the maximisation problem to derive the best-response function of each firm. Show the
two best-response functions in a graph that has p1 on the horizontal axis and p2 on the vertical axis
(when plotting the graph assume that V1 = V2 and k = 1).

Solution:
The first order condition for this problem is
∂D1
D 1 + p1 = 0
∂p1
 
1 V 1 − V 2 − p1 + p 2 1
+ − p1 = 0
2 2k 2k
1 (V1 − V2 ) p2 p1
+ + − = 0
2 2k 2k k
Hence, solving for the best response p1 , yields
1
p1 (p2 ; V1 , V2 ) = [k + (V1 − V2 ) + p2 ]
2
By a similar calculation, we obtain that the best price-response by firm 2 is
1
p2 (p1 ; V1 , V2 ) = [k + (V2 − V1 ) + p1 ] .
2

Inspecting the best-response function p1 (p2 ; V1 , V2 ) , we see firm 1’s chosen price increases by £0.5
for every £1 increase in p2 . Hence in terms of a graph that has p1 and p2 on the horizontal and
vertical axis respectively, the slope of p1 (p2 ; V1 , V2 ) is 2. By the same argument, the slope of
p2 (p2 ; V1 , V2 ) is 1/2. The following graph illustrates the case where k = 1 and V1 = V2 .

Downloaded by AKSHITI KACHHAWAH (2020.akshiti.kachhawah@ves.ac.in)

 lOMoARcPSD|32983522

EC3324 - Topics in Game Theory Exam Practice Questions - Spring 2019

p1 (p2 )

2
p2 (p1 )

p2
1

0
0 0.5 1 1.5 2 2.5 3
p1

(d) 1 Mark Solve for the Nash equilibrium set of prices given that V1 = V2 .

Solution:
With V1 = V2 the best response functions simplify to
k p2
p1 (p2 ; V1 , V2 ) = +
2 2
k p1
+
p2 (p1 ; V1 , V2 ) =
2 2
and it is easy to see that the unique price equilibrium is the symmetric outcome

p1 = p2 = k.

Total Marks for Question 2: 15

Downloaded by AKSHITI KACHHAWAH (2020.akshiti.kachhawah@ves.ac.in)

 lOMoARcPSD|32983522

EC3324 - Topics in Game Theory Exam Practice Questions - Spring 2019

Problem 3: Signaling Game

A Coach offers to help Alex prepare for his upcoming golf tournament. If Alex does not hire the coach, then
he will lose the tournament. If he hires the coach, then the probability he wins the tournament depends on
the quality of the coach. Specifically:
• If he hires a “good” coach, he will win with probability 3/4.
• If he hires a “bad” coach, he will win with probability 1/4.
Assume that Alex receives the equivalent of a monetary prize V = 100 if he wins the tournament and V = 0 if
he loses. Alex seeks to maximise his expected payoff net of any fees. So, if he pays x to the coach and expects
to win the tournament with probability q, then his expected payoff is given by: Es = V · q − x = 100q − x.
The coach seeks to maximise his expected payoffs from the coaching. The coach’s payoff is simply given by
x if he is hired and 0 if he is not hired.
The timing of events is as follows: the coach first sets a price x, Alex observes the price then decides whether
to hire the coach or not. Assume that both players only use pure strategies.

Please answer the following questions.

(a) 2 Marks Suppose that Alex knows he is playing against a “good” coach -
(i.e. P (coach=good) = 1), what price x would the “good” coach offer to Alex?

Solution:
Since the coach is “good”, Alex’s expected value of hiring the coach is given by:
3
Es = · 100 − x
4
If Alex doesn’t hire the coach he will lose the tournament and get a payoff of 0. Therefore Alex
will hire the coach as long as:
3
· 100 − x ≥ 0 ⇒ x ≤ 75
4
The “good” coach prefers to get hired as long as x ≥ 0. Since Alex is willing to pay up to 75 for a
“good” coach, the “good” coach will choose:

x = 75

(b) 5 Marks Suppose now that Alex believes a good or bad coach is equally likely (i.e. P (coach=good) =
1/2). In addition, the coach can only choose between two prices x1 and x2 where x1 > x2 and x1 ≤ 50.
Represent the game using an extensive form representation.

Solution:

Downloaded by AKSHITI KACHHAWAH (2020.akshiti.kachhawah@ves.ac.in)

 lOMoARcPSD|32983522

EC3324 - Topics in Game Theory Exam Practice Questions - Spring 2019

(x1 , 75 − x1 ) Coach (x2 , 75 − x2 )

H H
t = good
x1 x2

[α] [β]
(0, 0) NH NH (0, 0)
1/2

Alex N Alex

(x1 , 25 − x1 ) 1/2 (x2 , 25 − x2 )

H H
x1 x2

Coach
(0, 0) NH t = bad NH (0, 0)

(c) 5 Marks Is there Perfect Bayesian (separating) equilibrium in which both types of coach are hired by
Alex? If yes, state the strategy profile and the associated beliefs. If no, explain why there is not.

Solution:
There are two cases to consider:

• (i) The “good” coach chooses x1 and the “bad” coach chooses x2 .
• (ii) The “good” coach chooses x2 and the “bad” coach chooses x1 .

In the first case the extensive form is:

(x1 , 75 − x1 ) Coach (x2 , 75 − x2 )
H H
t = good
x1 x2

[α] [β]
(0, 0) NH NH (0, 0)
1/2

Alex N Alex

(x1 , 25 − x1 ) 1/2 (x2 , 25 − x2 )

H H
x1 x2

Coach
(0, 0) NH t = bad NH (0, 0)

Clearly this is not a PBE because the “bad” coach has a profitable deviation - specifically the “bad”
coach would prefer to deviate and choose x1 since x1 > x2 .

For the second case the extensive form is:

Downloaded by AKSHITI KACHHAWAH (2020.akshiti.kachhawah@ves.ac.in)

 lOMoARcPSD|32983522

EC3324 - Topics in Game Theory Exam Practice Questions - Spring 2019

(x1 , 75 − x1 ) Coach (x2 , 75 − x2 )

H H
t = good
x1 x2

[α] [β]
(0, 0) NH NH (0, 0)
1/2

Alex N Alex

(x1 , 25 − x1 ) 1/2 (x2 , 25 − x2 )

H H
x1 x2

Coach
(0, 0) NH t = bad NH (0, 0)

Clearly this is not a PBE because the “good” coach now has a profitable deviation - specifically
the “good” coach would prefer to deviate and choose x1 since x1 > x2 .

Therefore there is not a Perfect Bayesian (separating) equilibrium where both types of coach get
hired by Alex.

(d) 13 Marks Is there Perfect Bayesian (pooling) equilibrium in which both types of coach choose x1 and
are hired by Alex? If yes, state the strategy profile and the associated beliefs. If no, explain why there
is not.

Solution:
The extensive form for this strategy profile is:
(x1 , 75 − x1 ) Coach (x2 , 75 − x2 )
H H
t = good
x1 x2

[α] [β]
(0, 0) NH NH (0, 0)
1/2

Alex N Alex

(x1 , 25 − x1 ) 1/2 (x2 , 25 − x2 )

H H
x1 x2

Coach
(0, 0) NH t = bad NH (0, 0)

Clearly if both types of coach offer x1 and get hired, this is better than offering x2 irrespective of
Alex’s choice at that information set since x1 > x2 and x1 , x2 ≥ 0.

We need to check that when both types of coach offer x1 , hiring the coach is a best response for
Alex. The expected payoff for Alex from hiring the coach is:
Ea [H|x1 ] = α(75 − x1 ) + (1 − α)(25 − x1 ) = 50α + 25 − x1
The expected payoff from not hiring the coach is:
Ea [N H|x1 ] = 0

Downloaded by AKSHITI KACHHAWAH (2020.akshiti.kachhawah@ves.ac.in)

 lOMoARcPSD|32983522

EC3324 - Topics in Game Theory Exam Practice Questions - Spring 2019

Therefore Alex is willing to hire the coach iff:

50α + 25 − x1 ≥ 0

When both types of coach choose x1 Alex’s belief α is (using Bayes’ rule):

1/2
α= = 1/2
1/2 + 1/2

Which means that Alex will hire the coach iff:

50α + 25 − x1 ≥ 0 = 50(1/2) + 25 − x1 ≥ 0 ⇒ x1 ≤ 50

Which is true by definition. Therefore when both types of coach offer x1 , it is a best response for
Alex to hire the coach.

The last thing we need to check is what happens at the Alex’s other information set. Alex can
either choose H or N H since it will not affect either type of coachs’ choice. However, whatever we
specify for Alex at this information set has to be consistent with his beliefs - β.
If Alex chooses H, the extensive form is:
(x1 , 7.5 − x1 ) Coach (x2 , 7.5 − x2 )
H H
t = good
x1 x2

[1/2] [β]
(0, 0) NH NH (0, 0)
1/2

Alex N Alex

(x1 , 2.5 − x1 ) 1/2 (x2 , 2.5 − x2 )

H H
x1 x2

Coach
(0, 0) NH t = bad NH (0, 0)

This requires that:

Ea [H|x2 ] = β(75 − x2 ) + (1 − β)(25 − x2 ) = 50β + 25 − x2 ≥ 0 = Ea [N H|x2 ]

Which implies that:

x2 − 25
β≥
50

Therefore we now have a Perfect Bayesian (pooling) equilibrium:

{(x1 , x1 ), (H, H)}

x2 −2.5
with α = 1/2 and β ≥ 5
If Alex chooses N H, the extensive form is:

Downloaded by AKSHITI KACHHAWAH (2020.akshiti.kachhawah@ves.ac.in)

 lOMoARcPSD|32983522

EC3324 - Topics in Game Theory Exam Practice Questions - Spring 2019

(x1 , 75 − x1 ) Coach (x2 , 75 − x2 )

H H
t = good
x1 x2

[1/2] [β]
(0, 0) NH NH (0, 0)
1/2

Alex N Alex

(x1 , 25 − x1 ) 1/2 (x2 , 25 − x2 )

H H
x1 x2

Coach
(0, 0) NH t = bad NH (0, 0)

This requires that:

Ea [H|x2 ] = β(75 − x2 ) + (1 − β)(25 − x2 ) = 50β + 25 − x≤ 0 = Es [N H|x2 ]

Which implies that:

x2 − 25
β≤
50
Therefore we have another Perfect Bayesian (pooling) equilibrium:

{(x1 , x1 ), (H, N H)}

x2 −2.5
with α = 1/2 and β ≤ 5 .

In summary, there are two sets of Perfect Bayesian (pooling) equilibria in which both types of coach
choose x1 and are hired by Alex. These are given by:

{(x1 , x1 ), (H, H)}

x2 −2.5
with α = 1/2 and β ≥ 5

and

{(x1 , x1 ), (H, N H)}

x2 −2.5
with α = 1/2 and β ≤ 5

Total Marks for Question 3: 25

Downloaded by AKSHITI KACHHAWAH (2020.akshiti.kachhawah@ves.ac.in)

 lOMoARcPSD|32983522

EC3324 - Topics in Game Theory Exam Practice Questions - Spring 2019

Problem 4: Incomplete Information Game

Consider the Following game:

Where nature determines if payoffs are as in G1 or G2 by equal probability.

G1 G2
Player 2 Player 2
L C R L C R
T 2, 0 1, 1 4, 2 T 2, 0 1, 1 2, 0
Player 1 M 3, 4 1, 2 2, 3 Player 1 M 3, 4 1, 2 2, 3
B 1, 3 0, 2 3, 0 B 1, 3 0, 2 3, 0

Players move simultaneously. Player 2 knows which game nature has chosen, but player 1 does not. Please
answer the following questions.

(a) 5 Marks What is the strategy space for player 1? What about player 2?

Solution:
Remember that a strategy in a Bayesian game must specify an action for each type. Since player 1
does not know the game that will be played he can only specify one action. Player 2 however, can
specify an action for each realization of nature’s actions. Therefore the strategy sets are:

s1 = {T, M, B}

s2 = {(L, L), (L, C), (L, R), (C, L), (C, C), (C, R), (R, L), (R, C), (R, R)}

(b) 15 Marks Find all Bayesian Nash equilibria.

Solution:
Once we know the possible strategies for each player we can draw the normal form representation
and solve for the Bayesian Nash equilibrium. The normal form for this problem is:

Player 2
L, L L, C L, R C, L C, C C, R R, L R, C R, R
T 2, 0 1.5, 0.5 2, 0 1.5, 0.5 1, 1 1.5, 0.5 3, 1 2.5, 1.5 3, 1
Player 1 M 3, 4 2, 3 2.5, 3.5 2, 3 1, 2 1.5, 2.5 2.5, 3.5 1.5, 2.5 2, 3
B 1, 3 0.5, 2.5 2, 1.5 0.5, 2.5 0, 2 1.5, 1 2, 1.5 1.5, 1 3, 0

The way the payoffs in the above table are calculated is analogous to the method used in the
problem set. For example, when Player 1’s strategy is T , Player 2’s strategy is (L, L). Player 1’s

Downloaded by AKSHITI KACHHAWAH (2020.akshiti.kachhawah@ves.ac.in)

 lOMoARcPSD|32983522

EC3324 - Topics in Game Theory Exam Practice Questions - Spring 2019

expected payoff is:

1 1
E1 (s1 , s2 ) = ·2+ ·2=2
2 2

There are two Bayesian Nash equilibria in this game:

n o
• s∗ = (T ), (R, C)
n o
• s∗ = (M ), (L, L)

Total Marks for Question 4: 20

Downloaded by AKSHITI KACHHAWAH (2020.akshiti.kachhawah@ves.ac.in)

 lOMoARcPSD|32983522

EC3324 - Topics in Game Theory Exam Practice Questions - Spring 2019

Problem 5: Pure and Mixed Strategies

Consider the following normal form game:
Player 2
L M R
T 0, 8 5, 5 10, 2
Player 1
B 7, 1 2, 5 0, 8

Please answer the following questions:

(a) 2 Marks Consider only pure strategies. What does it mean if a strategy, s′i , strictly dominates another
strategy, si , for player i?

Solution:
A strategy s′i strictly dominates another strategy si for player i if:

ui (s′i , s−i ) > ui (si , s−i ) for all s−i ∈ S−i

(b) 2 Marks Consider only pure strategies. Does player 1 have a strictly dominated strategy? What
about player 2?

Solution:
Neither player has a pure strategy that is strictly dominated by another pure strategy.

(c) 2 Marks Is there a pure strategy Nash equilibrium in the game above?

Solution:
No there is no NE in pure strategies.

(d) 4 Marks Is there a mixed strategy Nash equilibrium in which player 1 plays a mixed strategy of the
form: σ1 = (p, 1 − p) and player 2 plays a mixed strategy of the form: σ2 = (q, 0, 1 − q)? If there is,
find the values of p ∈ (0, 1) and q ∈ (0, 1) in the equilibrium. If there is not, explain why it is not an
equilibrium.

Solution:
For the above strategy profiles to be a mixed strategy NE the following must be true:

E1 [T |σ2 ] = 10(1 − q) = 7q = E1 [B|σ2 ]

This implies that:

10 − 10q = 7q ⇒ q = 10/17

Downloaded by AKSHITI KACHHAWAH (2020.akshiti.kachhawah@ves.ac.in)

 lOMoARcPSD|32983522

EC3324 - Topics in Game Theory Exam Practice Questions - Spring 2019

We also require that:

E2 [L|σ1 ] = 8p + (1 − p) = 2p + 8(1 − p) = E1 [R|σ1 ]

This implies that:

1 + 7p = 8 − 6p ⇒ p = 7/13
Therefore there is a mixed strategy NE given by:

σ1 = (7/13, 6/13), σ2 = (10/17, 0, 7/17)

(e) 5 Marks Find all mixed strategies for player 2 that strictly dominate pure strategy M

Solution:
Define σ1 = (T − 1 − T ) and σ2 = (L, M, 1 − L − M ). If player 2 chooses pure strategy M her
expected payoff is:
E2 [M |σ1 ] = 5
Therefore we need to find all mixed strategies σ2 that give a greater expected payoff than 5.
The expected payoff from the mixed strategy σ2 is:

E2 [σ2 |σ1 ] = 8T L + 5T M + 2T (1 − L − M ) + (1 − T )L + 5(1 − T )M + 8(1 − T )(1 − L − M )

We require that this expected payoff is greater than 5 for all σ1 . This can be verified by setting
T = 0 and T = 1 and checking for dominance in both cases. In addition, it is clear that the mixed
strategy σ2 must assign M < 1 in order to strictly dominate. When T = 0 we have:

E2 [σ2 |σ1 ] = L + 5M + 8(1 − L − M ) > 5

solving for L gives:

3(1 − M )
L<
7
When T = 1 we have:
E2 [σ2 |σ1 ] = 8L + 5M + 2(1 − L − M ) > 5
Solving for L gives:
3(1 − M )
L>
6
Therefore, all mixed strategies σ2 for player 2 with:

1−M 3(1 − M )
<L< and M <1
2 7
strictly dominate pure strategy M .

Total Marks for Question 5: 15

Downloaded by AKSHITI KACHHAWAH (2020.akshiti.kachhawah@ves.ac.in)

 lOMoARcPSD|32983522

EC3324 - Topics in Game Theory Exam Practice Questions - Spring 2019

Problem 6: Three Player Normal Form Game

Consider the following normal form game where Player 1 chooses the row (either T or B), Player 2 chooses
the column (either l or r), and Player 3 chooses the table (either L or R).

Player 3
L R
Player 2 Player 2
l r l r
Player 1 T 2, 2, 2 0, 0, 0 T 0, 0, 0 0, 0, 0
B 0, 0, 0 0, 0, 0 B 0, 0, 0 5, 5, 5

Please answer the following questions:

(a) 2 Marks Find all the Nash equilibria in pure strategies.

Solution:

Player 3
L R
Player 2 Player 2
l r l r
Player 1 T 2, 2, 2 0, 0, 0 T 0, 0, 0 0, 0, 0
B 0, 0, 0 0, 0, 0 B 0, 0, 0 5, 5, 5

Finding the best responses for each player it is clear that there are two Nash equilibria in pure
strategies, these are:
{T, l, L}, {B, l, L}

(b) 3 Marks Assume that Player 1 moves first, then Player 2, and then Player 3. Moreover, all players
observe the choices of their predecessors. Represent this scenario using an extensive form representation.

Solution:
The extensive form representation for this game is shown below:

Downloaded by AKSHITI KACHHAWAH (2020.akshiti.kachhawah@ves.ac.in)

 lOMoARcPSD|32983522

EC3324 - Topics in Game Theory Exam Practice Questions - Spring 2019

P1

T B

P2 P2

l r l r

P3 P3 P3 P3

L R L R L R L R

               
2 0 0 0 0 0 0 5
2 0 0 0 0 0 0 5
2 0 0 0 0 0 0 5

(c) 10 Marks Using the extensive form representation in (b), find all sub-game perfect Nash equilibria.

Solution:
For an equilibrium to be sub-game perfect, requires that each player’s strategy specifies an action
at each information set that leads to a Nash equilibrium in every sub-game. To find such an equi-
librium, we first need to identify all sub-games.
As you can see, there are 7 sub-games in this game, we should start by solving for the Nash-
equilibrium in the sub-games at the bottom of the tree and then begin to work backwards up the
tree. The tree below highlights the best responses for each player in every sub-game (in red)).

P1

T B

P2 P2

l r l r

P3 P3 P3 P3

L R L R L R L R

               
2 0 0 0 0 0 0 5
2 0 0 0 0 0 0 5
2 0 0 0 0 0 0 5

Downloaded by AKSHITI KACHHAWAH (2020.akshiti.kachhawah@ves.ac.in)

 lOMoARcPSD|32983522

EC3324 - Topics in Game Theory Exam Practice Questions - Spring 2019

The backward induction outcome in this game is therefore:

Player 1 plays B, player 2 plays r, and player 3 plays R.

However there are 4 Nash equilibria that are sub-game perfect, these are:

• (B), (l, r), (L, L, L, R)

• (B), (l, r), (L, L, R, R)

• (B), (l, r), (L, R, L, R)

• (B), (l, r), (L, R, R, R)

Total Marks for Question 6: 15

Downloaded by AKSHITI KACHHAWAH (2020.akshiti.kachhawah@ves.ac.in)

 lOMoARcPSD|32983522

EC3324 - Topics in Game Theory Exam Practice Questions - Spring 2019

Problem 7: Perfect Bayesian Equilibrium (PBE)

P1

R L A

P2
(0, 4)
L′ R′ L′ R′

(−4, 0) (4, 2) (−2, −2) (2, 2)

(a) 5 Marks Find all Nash equilibria.

Solution:
We can draw the normal form representation for this game (below):

Player 2
L′ R′
L −4, 0 2, 2
Player 1 R −2, −2 4, 2
A 0, 4 0, 4

From this it is clear that there are two Nash equilibria:

• {(A), (L′ )},

• {(R), (R′ )}.

(b) 10 Marks For each Nash equilibria you found in part (a), check if they are also perfect Bayesian
equilibria (PBE).

Solution:
Consider equilibrium {(A), (L′ )}. For Player 2 to play L′ is never optimal (for all possible beliefs p it
is optimal for Player 2 to play R′ . Then this Nash equilibrium is not a perfect Bayesian equilibrium.

Now consider equilibrium {(R), (R′ )}. For Player 2 to play R′ is optimal for all possible beliefs p.
Given that Player 1 plays R, the value of p has to equal 0 (p = 0). Given Player 2’s strategy it is
optimal for Player 1 to play R.

Downloaded by AKSHITI KACHHAWAH (2020.akshiti.kachhawah@ves.ac.in)

 lOMoARcPSD|32983522

EC3324 - Topics in Game Theory Exam Practice Questions - Spring 2019

Then the strategy profile {(R), (R′ )} with p = 0 is a perfect Bayesian equilibrium.

Total Marks for Question 7: 15

Downloaded by AKSHITI KACHHAWAH (2020.akshiti.kachhawah@ves.ac.in)

 lOMoARcPSD|32983522

EC3324 - Topics in Game Theory Exam Practice Questions - Spring 2019

Problem 8: Cournot Duopoly

Consider a Cournot duopoly with discrete quantities. Let q1 denote Firm 1’s quantity and let q2 denote
Firm 2’s quantity. The inverse demand function for the firms’ good is given by:

P = 27 − 3Q

Both firms have zero marginal cost (MC=0) and face no fixed costs. Only three discrete quantities are
allowed: each firm i chooses:
qi ∈ {1, 2, 3}

We start by analyzing the one-shot Cournot game (i.e. when T = 1).

Please answer the following questions.

(a) 2 Marks Write each firm’s profit function πi as a function of the two quantities, q1 and q2 .

Solution:
πi = (27 − 3qi − 3qj )qi where i, j ∈ {1, 2}, i 6= j

(b) 3 Marks Fill out the normal-form game table for the simultaneous move quantity setting game played
by the two firms.

Solution:
Firm 2
1 2 3
1 21, 21 18, 36 15, 45
Firm 1
2 36, 18 30, 30 24, 36
3 45, 15 36, 24 27, 27

(c) 5 Marks Use the normal-form game table to find the Cournot-Nash equilibria of the game. What are
the quantities produced, price, and profits in each of the equilibria?

Solution:
The Nash equilibrium is:
q1∗ = q2∗ = 3
The price is:
P =9
The profits are:
π1∗ = π2∗ = 27

Now, assume that firms repeat the stage game above infinitely many times (T = ∞) and that both
firms discount their future profits by the same discount factor δ ∈ (0, 1). Given this infinite repetition,

Downloaded by AKSHITI KACHHAWAH (2020.akshiti.kachhawah@ves.ac.in)

 lOMoARcPSD|32983522

EC3324 - Topics in Game Theory Exam Practice Questions - Spring 2019

the firms may be able to earn higher profits than in the one-shot game by engaging in collusion. Let δ ∗
be the minimum discount factor that sustains collusion. Please answer the following questions.
(d) 2 Marks How can the two firms increase their profits, π1 , π2 , (relative to the Cournot equilibrium
above) by colluding in the infinitely repeated game? Define explicitly the “grim” trigger strategies that
the firms are required to follow in order to achieve this. (Note: The firms are still able to use only the
three prices listed above).

Solution:
The grim trigger strategy for firm i is given by:
(
q coll = 2 if qi,s = q coll = 2 for all s<t
qi,t =
q N E = 3 otherwise

(e) 6 Marks Determine the range of values for δ such that the collusive outcome is played in every period
of the infinitely repeated game (using the grim trigger strategy from the previous question).

Solution:
The condition is:
Πcoll ≥ Πdev
which is true when:
π coll πN E
≥ π dev + δ
1−δ 1−δ
solving gives:
π dev − π coll
δ≥
π dev − π N E
Which gives:
36 − 30 6
δ≥ = = 0.66
36 − 27 9

(f) 2 Marks Would collusion in this model be possible if the game were repeated a large, but finite number
of times? If yes, what would the firms strategies be? If not, why not?

Solution:
There is a unique NE in the stage game q1∗ = q2∗ = 3 therefore finite repetition cannot induce any
outcome other than NE in every period - Nash folk theorem. Therefore collusion would not be
possible in this case.

Total Marks for Question 8: 20

Downloaded by AKSHITI KACHHAWAH (2020.akshiti.kachhawah@ves.ac.in)

 lOMoARcPSD|32983522

EC3324 - Topics in Game Theory Exam Practice Questions - Spring 2019

Problem 9: Announcing Integers

The following game is player between two players: each player simultaneously announces an integer between
0 and 100. Let s1 be the announcement of player 1 and s2 be the announcement of player 2. Their payoffs
are determined as follows:
• If s1 + s2 ≤ 100: Player 1 receives s1 and Player 2 receives s2 ;
• If s1 + s2 > 100 and s1 > s2 : Player 1 receives 100 − s2 and Player 2 receives s2 ;
• If s1 + s2 > 100 and s1 < s2 : Player 1 receives s1 and Player 2 receives 100 − s1 ;
• If s1 + s2 > 100 and s1 = s2 : Both players receive 50.
Please answer the following questions:

(a) 1 Mark What is the set of pure strategies, Si , for each player i = 1, 2? What is the set of all pure
strategy profiles S?

Solution:
The set of pure strategies Si for each player is given by:

Si = [0, 100]

The set of all pure strategy profiles is the cartesian product of the two sets Si of pure strategies
si ∈ Si for i = 1, 2. In this case this is given by the following:

S = S1 × S2 = [0, 100]2

In this case the set of all pure strategy profiles is large so it is more appropriate to write it by
description (instead of enumeration).

(b) 9 Marks Find the solution(s) of this game by using the iterated elimination of strictly dominated
strategies.

Solution:
First notice that if player 1 announces s1 = 51, her payoff is given by:

• 50 if player 2 announces s2 = 50 or s2 = 51, and

• 51 if Player 2 announces anything else.

This observation is true for Player 2.

First round: If player 1 announces s1 < 51 she’ll get s1 irrespective of what Player 2 announces.
Therefore, any strategy smaller than 51 is dominated by 51, Likewise for Player 2. We can then
eliminate all strategies between 0 and 50 for both players.

Second round: If player 1 announces 100 she can get at most 50. This is because Player 2
announces a number between 51-100. Therefore 100 is dominated by 51 in this reduced game.
Likewise for Player 2. We can then eliminate 100 for both players.

Downloaded by AKSHITI KACHHAWAH (2020.akshiti.kachhawah@ves.ac.in)

 lOMoARcPSD|32983522

EC3324 - Topics in Game Theory Exam Practice Questions - Spring 2019

Third round: If player 1 announces 99 she can get at most 50. This is because Player 2 announces
a number between 51-99. Therefore 99 is dominated by 51 in this further reduced game. Likewise
for Player 2. We can then eliminate 99 for both players.

etc ....

Fourty-ninth round: If player 1 announces 53 she can get at most 50. This is because Player 2
announces a number between 51-53. Therefore 53 is dominated by 51 in this further reduced game.
Likewise for player 2. We can then eliminate 53 for both players.

Fiftieth round: If player 1 announces 52 she can get at most 50. This is because Player 2 an-
nounces a number between 51-52. Therefore 52 is dominated by 51 in this further reduced game.
Likewise for player 2. We can then eliminate 52 for both players.

Hence, only 51 survives the iterated elimination of strictly dominated strategies for both players.
As a result, the payoff for each player is 50. Formally, the strategy profile for the solution of this
game is given by: s = {51, 51}.

Total Marks for Question 9: 10

Downloaded by AKSHITI KACHHAWAH (2020.akshiti.kachhawah@ves.ac.in)

 lOMoARcPSD|32983522

EC3324 - Topics in Game Theory Exam Practice Questions - Spring 2019

Problem 10: Extensive Form Game

Consider the following extensive form game:

P1

T B

P2 P2

l r l r

P3 P1 P3 P3

L R L R L R L R

               
1 0 0 4 2 0 0 0
0 0 1 4 1 5 3 4
2 7 0 4 1 0 2 0

Please answer the following questions.

(a) 2 Marks Is this a game of perfect or imperfect information?

Solution:
Imperfect information.

(b) 2 Marks How many information sets does each player have?

Solution:
Player 1 → 2
Player 2 → 2
Player 3 → 2

(c) 2 Marks How many strategies does each player have?

Solution:
Player 1 → 4
Player 2 → 4
Player 3 → 4

(d) 3 Marks How many sub-games are there?

Downloaded by AKSHITI KACHHAWAH (2020.akshiti.kachhawah@ves.ac.in)

 lOMoARcPSD|32983522

EC3324 - Topics in Game Theory Exam Practice Questions - Spring 2019

Solution:
In total there are 5 subgames.

(e) 6 Marks What is the sub-game perfect Nash equilibrium (SPNE) in this game?

Solution:
Solve by backward induction starting from subgames at the end of the extensive form.
P1

T B

P2 P2

l r l r

P3 P1 P3 P3

L R L R L R L R

               
1 0 0 4 2 0 0 0
0 0 1 4 1 5 3 4
2 7 0 4 1 0 2 0

There is a unique SPNE, given by:

{(T, R), (r, r), (R, L)}

Total Marks for Question 10: 15

Downloaded by AKSHITI KACHHAWAH (2020.akshiti.kachhawah@ves.ac.in)