UNIVERSITY CARLOS III

Master in Economics Master in Industrial Economics and Markets

Game Theory
FINAL EXAM–January 19th, 2022

NAME:

Problem 1: Consider a Cournot duopoly with two firms which operate in a Market with one good and inverse
demand function P = 108 − Q where Q is the total output in the market. Firm 1 has a constant marginal cost
c1 = 27. Firm 2 has a constant marginal cost which can be either cl = 18 (with probability 1/3) or ch = 36 (with
probability 2/3). Firm 2 knows its cost and the cost of firm 1. Firm 1 knows its cost but the only information it has
about the cost of firm 2 is the above probability.

(a) (5 points.) Describe the situation as a Bayesian game

Solution: There are two players N = {1, 2}, There is one type of player for player 1, T1 = {c} and two types
for player 2 T2 = {l, h}. The beliefs of the players are

µ1 (t2 = l|t1 = c) = 1/3 µ1 (t2 = h|t1 = c) = 2/3

µ1 (t1 = c|t2 = l) = µ1 (t1 = c|t2 = h) = 1

The strategy sets are S1 = Sl = Sh = [0, +∞). The utilities of the players are
1 2
u1 (q1 , ql , qh ; c) = (108 − q1 − ql ) q1 + (108 − q1 − qh ) q1 − 27q1 =
3 3
ul (q1 , ql ) = u2 (q1 , ql , ql ; l) = (90 − q1 − ql ) ql = 90ql − q1 ql − ql2
uh (q1 , qh ) = u2 (q1 , ql , qh ; h) = (72 − q1 − qh ) qh = 72qh − q1 qh − qh2

(b) (10 points.) Compute the best reply of each firm. You must compute the best reply of each type of the firms.

Solution: Agent 2, type h, maximizes maxqh uh = (72 − qh − q1 )qh . The first order condition is

72 − 2qh − q1 = 0.

Note that the second derivative with respect to qh is

∂ 2 uh
= −2 < 0
∂qh2
Hence, the first order condition corresponds to a maximum of uh . The best reply of agent 2, type h, is
72 − q1
BRh (q1 ) =
2

Likewise, agent 2, type l, maximizes maxql ul = (90 − 2ql h − q1 )ql . The first order condition is

90 − 2ql − q1 = 0.

Note that the second derivative with respect to ql is

∂ 2 ul
= −2 < 0
∂ql2

1
 Hence, the first order condition corresponds to a maximum of ul . The best reply of agent 2, type l, is
90 − q1
BRl (q1 ) =
2

Finally, agent 1 maximizes

1 2
max (108 − q1 − ql ) q1 + (108 − q1 − qh ) q1 − 27q1
q1 3 3
The first order condition is
1 2
81 − 2q1 − q1 − qh = 0
3 3
The best reply of agent 1 is
1
BR1 (qh , ql ) = (243 − 2qh − ql )
6

(c) (10 points.) Compute the Bayes–Nash equillibrium, the quantities sold in this equilibrium and the profits of
each firm.

Solution: The NE is the solution to

72 − q1 90 − q1 1
qh = ql = q1 = (243 − 2qh − ql )
2 2 6
We obtain

qh∗ = 22 ql∗ = 31 q1∗ = 28

the utilities of the agents are
u∗h = 484 u∗l = 961 u∗1 = 784

(d) (10 points.) Suppose now that it is known that firm 2 has cost cl = 18. What is the Cournot equilibrium and
the profits of the firms? Which firm is better off and which one is worse off with respect to case of asymmetric
information?

Solution: The best reply of agent 2, type l, is still

90 − q1
BRl (q1 ) =
2
But now agent 1 maximizes
max (108 − q1 − ql ) q1 − 27q1
q1

The first order condition is

81 − 2q1 − qh = 0
The best reply of agent 1 is
81 − q2
BR1 (qh , ql ) =
2
The NE is the solution to
90 − q1 81 − q2
ql = q1 =
2 2
We obtain

2
 ql∗ = 33 q1∗ = 24
the utilities of the agents are
u∗l = 1089 u∗1 = 576
Firm 2 type l is better off in comparison with the asymmetric situation. Firm 1 is worse off.

(e) (10 points.) Suppose now that it is known that firm 2 has cost ch = 36. What is the Cournot equilibrium and
the profits of the firms? Which firm is better off and which one is worse off with respect to case of asymmetric
information?

Solution: The best reply of agent 2, type h, is

72 − q1
BRh (q1 ) =
2
And now agent 1 maximizes
max (108 − q1 − qh ) q1 − 27q1
q1

The first order condition is

81 − 2q1 − qh = 0
The best reply of agent 1 is
81 − q2
BR1 (qh , qh ) =
2
The NE is the solution to
72 − q1 81 − q2
qh = q1 =
2 2
We obtain

qh∗ = 21 q1∗ = 30
the utilities of the agents are
u∗h = 441 u∗1 = 900
Firm 2 type l is worse off in comparison with the asymmetric situation. Firm 1 is better off.

(f) (10 points.) Suppose now that firm 1 does not know the cost of firm 2. But, firm 1 can hire a research insti-
tute that will publish the cost of firm 2. How much would firm 1 be willing to pay to make that information public?

Solution: Without the information the profit of firm 1 is 784. If it is common knowledge that firm 2 is of type
l, the profit of firm 1 would be 576. If it is common knowledge that firm 2 is of type h, the profit of firm 1 would
be 900. Thus, the expected value of revealing the information is
1 2
× 576 + × 900 = 792
3 3
Therefore, the maximum amount firm 1 would be willing to pay is 792 − 784 = 8.

Problem 2: Consider the situation in which player 1 knows what is the game played (a or b below). But, player 2
only knows that a is played with probability 1/2 and b is played with probability 1/2.

Player 2 Player 2
X Y X Y
Player 1 C 18, 8 0, 0 Player 1 C 18, 0 0, 12
D 0, 0 12 , 12 D 0, 8 12 , 0
(a, 1/2) (b, 1/2)

3
(10 points.) Find the Bayesian equilibria in pure strategies and the payoffs of the players in those equilibria.
Solution: Note that BR1 (X) = CC and
u2 (CC, X) = 4
u2 (CC, Y ) = 6
Hence, BR2 (CC) = Y . We conclude that there is no pure strategy NE in which player 2 plays X. On the other
hand, BR1 (Y ) = DD and
u2 (DD, X) = 4
u2 (DD, Y ) = 6
Hence, BR2 (DD) = Y . We conclude that (DD, Y ) is a BNE in pure strategies. The payoffs are ua (DD, Y ) = 12,
ub (DD, Y ) = 12. u2 (DD, Y ) = 6.

(20 points.) Find the Bayesian equilibria in mixed strategies and the payoffs of the players in those equilibria.
Solution: Let us look for a BNE in mixed strategies of the form
σa = xC + (1 − x)D, 0≤x≤1
σb = yC + (1 − y)D 0≤y≤1
σ2 = zX + (1 − z)Y 0≤z≤1

Suppose first that 0 < x < 1. Then ua (C, σ2 ) = ua (D, σ2 ), Since,

ua (C, σ2 ) = 18z
ua (D, σ2 ) = 12 − 12z

we must have z = 2/5. Given this value of z, we have

36
ub (C, σ2 ) = 18z =
5
36
ub (D, σ2 ) = 12 − 12z =
5

and player b is also indifferent between C and D for z = 2/5. Now, for z = 2/5, player 2 must be indifferent between
X and Y .
1 1
u2 (σa , σb ; X) = 18x + 8(1 − y) = 4 + 4x − 4y
2 2
1 1
u2 (σa , σb ; Y ) = 18(1 − x) + 12y = 6 − 6x + 6y
2 2

equating 4 + 4x − 4y = 6 − 6x + 6y we obtain 5x − 5y = 1. Since, x, y ≥ 0 we obtain a family of BNE

       
1 6 2 3 1
xC + (1 − x)D, x − C+ −x D , X + Y : ≤x≤1
5 5 5 5 5
with payoffs
36 24
ua = ub = , u2 = ,
5 5

Problem 3:
Consider the following game in extensive form

4
 1

Z W
1.1 1.2

A B C D
2.1 2.2

U V U V X Y X Y

0, 1 −2, 0 2, 2 10, 1 3, 4 1, 2 1, 5 5, 6

1. (5 points.) What are the sub-games of the above game? It is enough to write the node at which each sub-game
starts.
Solution: There are three sub-games: The whole game and the sub-games that start at the nodes 1.1 and 1.2.

2. (5 points.) Write the normal form of the sub-game that starts at at node 1.1. Find the Nash equilibria in
pure strategies of this sub-game.
Solution: The normal form of the sub-game that starts at at node 1.1 is,

U V
A 0, 1 −2, 0
B 2, 2 10, 1

There is a unique NE in pure strategies: (B, U ) with payoffs (2, 2).

3. (5 points.) Write the normal form of the sub-game that starts at at node 1.2. Find the Nash equilibria in
pure strategies of this sub-game.
Solution: The normal form of the sub-game that starts at at node 1.2 is,

X Y
C 3, 4 1, 2
D 1, 5 5, 6

There are two NE in pure strategies: (C, X) with payoffs (3, 4) and (D, Y ) with payoffs (5, 6).

4. (10 points.) Find the pure strategy sub-game perfect Nash equilibria of the complete game.
Solution: Let us write an strategy as (1, (1.1, 2.1), (1.2, 2.2)). At node 1, player 1 chooses Z and W anticipating
the subsequent NE which gives him the highest payoff. Thus, the sub-game perfect Nash equilibria are:

(a) (W, (B, U ), (C, X)), with payoffs (3, 4).

(b) (W, (B, U ), (D, Y ))), with payoffs (5, 6).

5
Problem 4: Consider the following stage game

C D
C 10, 10 0, 15
D 15, 0 5, 5

(a) (5 points.) Find all the NE of the above game.

Solution: The strategy C is strictly dominated for player 1. Once we eliminate this stragegy, C is strictly
dominated for player 2. Hence, the unique NE is (D, D) with payoff u1 = u2 = 5.

Consider now the repeated game which consists in playing the above stage game infinitely many times with discount
factor δ.

(b) (5 points.) Can you find a subgame perfect Nash Equilibrium for every 0 < δ < 1?

Solution: The strategy (D, D) is a NE of the stage game. Thus, playing (D, D) in every period is a SPNE.

(c) (10 points.) Describe the trigger strategy.

(d) (15 points.) For what values of δ does the trigger strategy constitute a subgame perfect Nash equilibrium?
You must justify why the trigger strategy is a SPNE for those values of δ.

Solution:
If players follow the trigger strategy their stream of payoffs is
10
10 + 10δ + 10δ 2 + · · · =
1−δ
If one player deviates at the first period his stream of payoffs will be
5δ
15 + 5δ + 5δ 2 + · · · = 15 +
1−δ
The trigger strategy is a NE of the repeated game iff
10 5δ
≥ 15 +
1−δ 1−δ

that is iff δ ≥ 21 . The standard argument shows it is a SPNE.

(d) Suppose now that δ = 3/4. Let k pe a positive integer. Consider the following ‘soft’ trigger stragegy:
• At t = 1 play (C, C).
• If 2 ≤ t ≤ k play (C, C) if (C, C) was played at t − 1, t − 2, . . . , 1. Play (D, D) otherwise.
• If t > k play (C, C) if (C, C) was played at t − 1, t − 2, . . . , t − k. Play (D, D) otherwise.
That is, if one player deviates from (C, C) he is punished for k periods with (D, D) and then we go back to
(C, C). What is the smallest value of k for which the above strategy is a sub-game perfect Nash equilibrium?
(15 points.) Solution: If players follow the ‘soft’ trigger strategy their stream of payoffs is
10
ut = 10 + 10δ + 10δ 2 + · · · =
1−δ

6
If one player deviates at the first period his stream of payoffs will be

ud = 15 + 5δ + 5δ 2 + · · · + 5δ k + 10δ k+1 + 10δ k+2 + · · ·

The ‘soft’ trigger strategy is a NE of the repeated game iff

ut − ud = −5 + 5δ + 5δ 2 + · · · + 5δ k ≥ 0

Let
x = 5δ + 5δ 2 + · · · + 5δ k
Then
δx = 5δ 2 + 5δ 3 + · · · + 5δ k+1
So,
(1 − δ)x = 5δ − 5δ k+1
Hence,
5δ − 5δ k+1
x=
1−δ
That is,
δ − δ k+1
5δ + 5δ 2 + · · · + 5δ k = 5
1−δ
Therefore, the ‘soft’ trigger strategy is a NE of the repeated game iff

δ − δ k+1
≥1
1−δ
That is, if
δ − δ k+1 ≥ 1 − δ
or
2δ ≥ 1 + δ k+1
3
with δ = 4 the above inequality becomes
 k+1
3 3
1+ ≤
4 2
For k = 1 we have  2
3 25 3
1+ = >
4 16 2
For k = 2 we have  3
3 91 3
1+ = <
4 64 2
So, it is enough to take k = 2.