UNIVERSITY CARLOS III

Master in Industrial Economics and Markets

Game Theory
Final Exam January 27, 2017
Solutions

Problem 1: Consider the following game in normal form.

A2 B2 C2 D2
A1 6,-1 3,3 4,0 2,2
B1 2, 301 2,-10 2,15 1,2
C1 5,-1 5,0 3,1 4,1

(1) (10 points) Which are the strategies that survive the iterated deletion of strictly dom-
inated strategies?

Solution: The stratey B1 is strictly dominated by A1 (and C1 ). After we eliminate

this strategy we obtain the game
A2 B2 C2 D2
A1 6,-1 3,3 4,0 2,2
C1 5,-1 5,0 3,1 4,1
Now, the strategy A2 e is strictly dominated by the strategies B2 , C2 and D2 . DAfter
we eliminate this strategy we obtain the following game
B2 C2 D2
A1 3,3 4,0 2,2
C1 5,0 3,1 4,1
Hence, the strategies that survive the iterated elimination of dominated strategies are
S1 = {A1 , C1 }, S2 = {B2 , C2 , D2 }.
(2) (5 points) Find all the Nash equilibria in pure strategies and the payoffs of the players.

Solution: The best responses of the players are

B2 C2 D2
A1 3,3 4,0 2,2
C1 5,0 3,1 4,1
The unique Nash equilibria in pure strategies is (C1 , D2 ). The payoffs are u1 (C1 , D2 ) = 4,
u2 (C1 , D2 ) = 1.

(3) (10 points) Find all the Nash equilibria in mixed strategies and the payoffs of the play-
ers.

Solution: Suppose that player 1 follows the mixed strategy s1 = pA1 + (1 − p)C1 and
that player 2 follows the mixed strategy s2 = q1 B2 + q2 C2 + (1 − q1 − q2 )D2 . The payoffs
for player 1 are

u1 (A1 , s2 ) = 3q1 + 4q2 + 2(1 − q1 − q2 ) = 2 + q1 + 2q2

u1 (C1 , s2 ) = 5q1 + 3q2 + 4(1 − q1 − q2 ) = 4 + q1 − q2
1
2

and the payoffs for player 2 are

u2 (s1 , B2 ) = 3p
u2 (s1 , C2 ) = 1 − p
u2 (s1 , D2 ) = 1 + p

Note first that 2 + q1 + 2q2 ≥ 4 + q1 − q2 iff q2 ≥ 32 . It follows that


2
C1 , if q2 < 3

BR1 (q1 B2 + q2 C2 + (1 − q1 − q2 )D2 ) = {A1 , C1 }, if q2 = 2
3
A1 , if q2 > 23



In particular, it follows that player 1 follows a mixed strategy only if q2 = 23 . On the

other hand, since 1 + p > 1 − p for p > 0 we see that player 2 will not use strategy C2
unless p = 0.
Can player 2 follow a mixed strategy with q2 = 0 and 0 < q1 < 1? In such a case,
we must have that u2 (s1 , B2 ) = u2 (s1 , D2 ) and we obtain that p = 21 . This means that
player 1 is using a completely mixed strategy. But, as we have proved above this is
possible only if q2 = 23 , which contradicts that q2 = 0. Hence, there are no NE with
q2 = 0 and 0 < q1 < 1.
Can player 2 follow a mixed strategy with q1 = 0 and 0 < q2 < 1? In such a case, we
must have that u2 (s1 , C2 ) = u2 (s1 , D2 ). That is, 1 − p = 1 + p which implies p = 0. This
means player 1 best response is C1 , so q2 ≤ 23 . Thus, we find that there are infinitely
many NE of the form

(C1 , q2 C2 + (1 − q2 )D2 )

with 0 ≤ q2 ≤ 32 . The payoffs are u1 = 4 − q2 , u2 = 1.

Problem 2: Agent 001 has to decide between completing his mission (M) and escorting the
president of the country, or to pretend (P) that he is ill and go to the town T where he would
play cards with agent 002. If agent 001 fulfills his duty, then agent 002 watches agent 001 on TV
protecting the president and would know that they are not going to play cards. In this situation
the utilities of the agents would be u1 = 2 and u2 = 0.
If agent 001 pretends to be ill then he would travel the the town T . In there, one can find
two bars where it is allowed to play cards: A and B. If both agents go to bar A their utilities
are u1 = 1 and u2 = 4. Whereas, if they both go to B their utilities are u1 = 4 and u2 = 1. If
they do not go to the same bar, their utilities would be 0 for both of them.

(1) Assume first that agent 001 cannot communicate with agent 002 to coordinate, because
the message would be intercepted by some hackers that would leak the fraud.
(a) (5 points) Describe the situation as a dynamic game. How many subgames does
this game have? Solution:
 3

1.1

M P
2.1
2, 0

A B
1.2

A B A B

1, 4 0, 0 0, 0 4, 1
There are two subgames. They start at the nodes 1.1 and 2.1.

(b) (5 points) Compute the Nash equilibria of all the proper subgames of the above
game and the payoffs of the players.

Solution: There is only one proper subgame. It starts at node 2.1. We write this
game in normal form
A B
A 1,4 0,0
C 0,0 4,1
There are two NE in pure strategies,
• (A, A) with payoffs u1 (A, A) = 1, u2 (A, A) = 4; and
• (B, B) with payoffs u1 (B, B) = 4, u2 (B, B) = 1.

Let us now search for a mixed strategy Ne of the form

(xA + (1 − x)B, yA + (1 − y)B)
Note that
u1 (A, yA + (1 − y)B) = y
u1 (B, yA + (1 − y)B) = 4 − 4y
u2 (xA + (1 − x)B, A) = 4x
u2 (xA + (1 − x)B, B) = 1−x

from where we obtain that x = 51 , y = 45 . So,

 
1 4 4 1
A + B, A + B
5 5 5 5
is a NE with payoffs
   
1 4 4 1 1 4 4 1 4
u1 A + B, A + B = u2 A + B, A + B =
5 5 5 5 5 5 5 5 5

(c) (5 points) Compute the subgame perfect Nash equilibria of the above game and
the payoffs of the players.

Solution: There are three SPNE

• ((M, A), A) with payoffs u1 = 2, u2 = 0;
4

• ((P, B), B) with payoffs u1 = 4, u2 (B, B) = 1; and

M, 51 A + 45 B , 54 A + 15 B with payoffs u1 = 2, u2 = 0.



•

(2) Assume now that agent 002 is able to leave a message for agent 001 at the train station
of town T with the bar that agent 2 will go to.
(a) (5 points) Describe the new situation as a dynamic game. How many subgames
does this game have?

Solution:
1.1

M P
2.1
2, 0

A B
1.2 1.3

A B A B

1, 4 0, 0 0, 0 4, 1
There are four subgames. They start at the nodes 1.1 and 2.1, 1.2 and 1.3.

(b) (5 points) Compute the subgame perfect Nash equilibria of this new game and the
payoffs of the players.
Solution:
1.1

M P
2.1
2, 0

A B
1.2 1.3

A B A B

1, 4 0, 0 0, 0 4, 1
The SPNE is ((M, A, B), A). The payoffs are u1 = 2, u2 = 0.

Problem 3: Consider the following game,with two players, in normal form.

Player 2
C D
Player 1 A 5,4 1,5
B 6, 2 3,3

and consider the repeated game with the above game as a stage game.
 5

(1) (5 points) Suppose that the above game is played two times. How many information
sets does each player have?, How many strategies does each player have? Give an exam-
ple of a possible strategy for player 1. How many subgames are in the game repeated
two times?

Solution:
• At the first stage each player has one information set. There are four possible
outcomes in the game. Hence, at the second stage each player has four information
sets. Therefore, each player has a total of five information sets.
• There are also 5 subgames; and
• Each player can choose among 2 strategies in every subgame. Hence, there is a total
of 25 = 32 strategies per player.
• An example of a strategy for player 1 could be
(a) play A in the first period.
(b) In the second period play
– A if in the first period (A, C) was played.
– B if in the first period (A, D) was played.
– A if in the first period (B, C) was played.
– B if in the first period (B, D) was played.

(2) (5 points) Suppose that the above game is played 2 times. Compute all the subgame
perfect Nash equilibria and the payoffs of the players.

Solution: The game has a unique NE: (B, D), with payoffs u1 (B, D) = u2 (B, D) = 3.
Hence, the game repeated finitely many times has a unique SPNE which consists in play-
ing the unique NE of the stage game at every period. Since it is repeated two times, the
payoffs of the players are u1 = u2 = 6.

(3) (5 points) Suppose that the above game is played 198743926508 times. Compute all
the subgame perfect Nash equilibria and the payoffs of the players.

Solution: By the argument above the payoffs are u1 = u2 = 3 × 198743926508.

(4) (10 points) Suppose that the above game is played infinitely many times and both
players have the same discount factor δ. Compute the smallest δ such that there is a
subgame perfect Nash equilibria, using trigger strategies, in which at every stage players
play the strategy (A, C).

Solution: The trigger strategy for player 1 is the following,

(a) At period t = 1 play A.
(b) If in every period 1 ≤ s < t it was played (A, C), then in period t play A. Otherwise,
play B.
And the trigger strategy player 2 is
(a) At period t = 1 play C.
(b) If in every period 1 ≤ s < t it was played (A, C), then in period t play C. Otherwise,
play D.
If both players follow the trigger strategy, the payoffs are

u1 = 5 + 5δ + 5δ 2 + 5δ 3 + · · ·
u2 = 4 + 4δ + 4δ 2 + 4δ 3 + · · ·
6

It is enough to consider deviations in the first period. If player 1 deviates in period 1,

and player 2 follows the trigger strategy the payoff of player 1 is
ud1 = 6 + 3δ + 3δ 2 + 3δ 3 + · · ·
Hence, player 1 does not have incentives to deviate iff
5 + 5δ + 5δ 2 + 5δ 3 + · · · ≥ 6 + 3δ + 3δ 2 + 3δ 3 + · · ·
That is if
2δ
2δ + 2δ 2 + 2δ 3 + · · · = ≥1
1−δ
which is equivalent to δ ≥ 13 .
If player 2 deviates in period 1, and player 1 follows the trigger strategy the payoff of
player 2 is
ud2 = 5 + 3δ + 3δ 2 + 3δ 3 + · · ·
Hence, player 2 does not have incentives to deviate iff
4 + 4δ + 4δ 2 + 4δ 3 + · · · ≥ 5 + 3δ + 3δ 2 + 3δ 3 + · · ·
That is if
δ
δ + δ2 + δ3 + · · · = ≥1
1−δ
which is equivalent to δ ≥ 12 . So, we need δ ≥ max{ 21 , 13 } = 12 .

Problem 4: Consider the situation in which player 2 knows which game is played (a or b
below). However, player 1 only knows that table a is played with probability 21 and table b is
played with probability 21 .

Player 2 Player 2
C D C D
Player 1 A 7,3 1,2 Player 1 A 7,1 1,6
B 2, 1 4,6 B 2, 3 4,2
a b
(1) (5 points) Describe the situation as a Bayesian game.

Solution: The set of players is N = {1, 2}. The set of strategies is S1 = {A, B},
S2 = {CC, CD, DC, DD}. The types are T1 = {c}, T2 = {a, b}. The beliefs are
1
p1 (a|c) = p1 (b|c) =
2
p2 (c|a) = p2 (c|b) = 1
The payoffs are described in the above tables.

(2) (10 points) Find the Bayesian–Nash equilibria in pure strategies and the payoffs of the
players.

Solution: We remark that BR2 (A) = CD and BR2 (B) = DC. Also
1 1
u1 (A, CD) = ×7+ ×1=4
2 2
1 1
u1 (B, CD) = ×2+ ×4=3
2 2
 7

Hence, BR1 (CD) = A and we see that (A, CD) is a BNE with payoffs u1 (A, CD) = 4,
u2 (A, CD|a) = 3,u2 (A, CD|b) = 6. On the other hand,
1 1
u1 (A, DC) = ×1+ ×7=4
2 2
1 1
u1 (B, DC) = ×4+ ×2=3
2 2
Hence, BR1 (DC) = A and we see that there are no other BNE in pure strategies.

(3) (10 points) Find the Bayesian–Nash equilibria in mixed strategies and the payoffs of
the players.

Solution: Let us look for a BNE of the form

(xA + (1 − x)B, (yC + (1 − y)D, zC + (1 − z)D))
Let
s1 = xA + (1 − x)B
sa = yC + (1 − y)D
sb = zC + (1 − z)D
We have that
1 1
u1 (A; sa , sb ) = (7y + 1 − y) + (7z + 1 − z) = 1 + 3y + 3z
2 2
1 1
u1 (B; sa , sb ) = (2y + 4(1 − y)) + (2z + 4(1 − z)) = 4 − y − z
2 2
u2 (s1 , C|a) = 3x + 1 − x = 1 + 2x
u2 (s1 , D|a) = 2x + 6(1 − x) = 6 − 4x
u2 (s1 , C|b) = x + 3(1 − x) = 3 − 2x
u2 (s1 , D|b) = 6x + 2(1 − x) = 2 + 4x

Suppose first that player 2a is using a completely mixed strategy. Then u2 (s1 , C|a) =
u2 (s1 , D|a). Hence, 1 + 2x = 6 − 4x and we conclude that x = 56 . For this value of x
we have that u2 (s1 , C|b) = (3 − 2x)|x= 5 = 43 and u2 (s1 , D|b) = (2 + 4x)|x= 5 = 16
3 , so
6 6
z = 0. We check if there is a BNE of the form
 
5 1
A + B, (yC + (1 − y)D, D)
6 6
Player 1 must be indifferent between A and B. Hence, 1 + 3y + 3z = 4 − y − z. Since
z = 0, we obtain that y = 43 . And we have checked that
  
5 1 3 1
A + B, C + D, D
6 6 4 4
is BNE in mixed strategies with payoffs u1 = 13 8 16
4 , u2 (·|a) = 3 , u2 (·|b) = 3 .
Suppose now that player 2b is using a completely mixed strategy. Then u2 (s1 , C|b) =
u2 (s1 , D|b). Hence, 3 − 2x = 2 + 4x and we conclude that x = 61 . For this value of x
we have that u2 (s1 , C|a) = (1 + 2x)|x= 1 = 43 and u2 (s1 , D|a) = (6 − 4x)|x= 1 = 16 3 , so
6 6
y = 0. We check if there is a BNE of the form
 
1 5
A + B, (D, zC + (1 − z)D)
6 6
8

Player 1 must be indifferent between A and B. Hence, 1 + 3y + 3z = 4 − y − z. Since

y = 0, we obtain that z = 43 . And we have checked that
  
1 5 3 1
A + B, D, C + D
6 6 4 4
13 16
is the other BNE in mixed strategies with payoffs u1 = 4 , u2 (·|a) = 3 , u2 (·|b) =
8
3.