Microeconomics: Decision Theory, MQuEA Prof.

Arguedas

Problem set #3
_____________________________________________________________________________

1. In the following incomplete information game, player 1 knows whether nature has drawn
Game A or Game B, but player 2 does not. The corresponding payoffs for the two players are
given in the following two matrices:

Game A L R
U 1, 1 0, 0
D 0, 0 0, 0

Game B L R
U 0, 0 0, 0
D 0, 0 2, 2

Compute:

(i) The pure-strategy Bayesian-Nash equilibria of the simultaneous version of this

game.
(ii) The Perfect Bayesian equilibrium of the sequential version of this game (assume
the informed player plays in the first place). Clearly outline the equilibrium beliefs
that sustain the equilibrium.
(iii) Explain the differences between the two types of equilibria found.

Solution:

(i)

In this game we can create an artificial matrix with the uniformed player 2 in the columns and
two sub players 1, one for each nature, in the rows:

L R
UA, UB 1, 0; 0.5 0, 0; 0
UA, DB 1, 0; 0.5 0, 2; 1
DA, UB 0, 0; 0 0, 0; 0
DA, DB 0, 0; 0 0, 2; 1
Microeconomics: Decision Theory, MQuEA Prof. Arguedas

Problem set #3
_____________________________________________________________________________

Notice that player 2 does not have any clue about the nature in which he will play, so he should
assign equal probabilities to nature A and nature B. Under this reasoning, he should calculate
his payoffs with probability=0.5, for example: (UA, UB; L)=( 1, 0; 0.5*1+0.5*0)=(1,0; 0.5).
Therefore, they can reach 3 different pure-strategy Bayesian-Nash equilibria: (UA, UB; L), (UA, DB;
R) and (DA, DB; R).

(ii)

Now we can use a decision tree and backwards induction:

(1,1)
L
2
(0,0)
R
U 1-q
L (0,0)
1 2
D R (0,0)
q
A
U 2 L (0,0)
1
R (0,0)
1-r
B (0,0)
D 2 L

r
R (2,2)

Where q and r reflect the beliefs of player 2 that player 1 does, respectively, U under nature B
and D under nature B; while (1-q) and (1-r) are the beliefs of player 2 that player 1 goes,
respectively, U under nature A and D under nature A. With this information, player 2 is analysed
first:

- If player 1 chooses U, player 2 goes for L if → 𝐸𝑈 (𝐿) ≥ 𝐸𝑈 (𝑅) ⇒ (1 − 𝑞) + 0𝑞 ≥

0(1 − 𝑞) + 0𝑞 ⇒ 𝑞 ≤ 1. Provided that 𝑞 ∈ (0,1], then 𝐸𝑈 (𝐿) ≥ 𝐸𝑈 (𝑅) is always
satisfied. Therefore, if player 1 chooses U, player 2 will always go for L.
- If player 1 chooses D, player 2 goes for L if → 𝐸𝑈 (𝐿) ≥ 𝐸𝑈 (𝑅) ⇒ 0(1 − 𝑟) + 0𝑟 ≥
0(1 − 𝑟) + 2𝑟 ⇒ 𝑟 ≤ 0. Provided that 𝑟 ∈ (0,1], then 𝐸𝑈 (𝐿) ≥ 𝐸𝑈 (𝑅) is never
satisfied. Therefore, if player 1 chooses D, player 2 will always go for R.

Recall why beliefs q and r cannot be equal to 0: A perfect Bayesian Nash equilibrium requires
“consistency on the path” (i.e. given any strategy profile 𝑠 and any information set 𝐼 on the path
Microeconomics: Decision Theory, MQuEA Prof. Arguedas

Problem set #3
_____________________________________________________________________________

of 𝑠, a player’s beliefs at 𝐼 is said to be consistent with 𝑠 if and only if the beliefs are derived
using the Bayes’ rule and 𝑠), and Bayes’ Rule cannot be computed for r or q equal 0 because the
denominator would be 0, which is undefined.

Knowing the previous, under nature A, player 1 will choose (U,L) over (D,L) because payoffs are
respectively (1,1) and (0,0). And under nature B, player 1 will choose (D,R) over (U,R) because
payoffs are respectively (2,2) and (0,0).

The Perfect Bayesian Equilibrium is (U,L) for nature A and (D,R) for nature B; for any q and any
r.

(iii)

In the first equilibrium, player 2 can only assume randomness regarding player 2 actions, thus
assigning p=1/2. But in the second equilibria, player 2 can infer nature from player 1 actions,
with two different actions of player 1 depending on the two natures, allowing player 2 to choose
the best response (i.e. separating equilibrium).

2. Consider the following two normal-form representations of a simultaneous game:

Game 1
D E
A 7, 6 2, 0
B 5, 8 1, 1
C 0, 0 4, 4
Game 2 D E
A 7, 6 0, 0
B 10, 8 1, 1
C 0, 5 3, 4

a) Assume player 1 knows whether the payoff matrix is that of game 1 or game 2, while
player 2 does not. Find all the pure Bayesian-Nash equilibria of the incomplete
information game.
b) Assume instead that players play sequentially. The informed player (player 1) moves
first. Then player 2 observes the action chosen by player 1 and moves afterwards.
Microeconomics: Decision Theory, MQuEA Prof. Arguedas

Problem set #3
_____________________________________________________________________________

Find the perfect Bayesian equilibrium of this game. Is the equilibrium pooling or
separating?
Solution
a)
D E
A1, A2 7, 7; 6 2, 0; 0
A1, B2 7, 10; 7 2, 1; 0.5
A1, C2 7, 0; 5.5 2, 3; 2
B1, A2 5, 7; 7 1, 0; 0.5
B1, B2 5, 10; 8 1, 1; 1
B1, C2 5, 0; 6.5 1, 3; 2.5
C1, A2 0, 7; 3 4, 0; 2
C1, B2 0, 10; 4 4, 1; 2.5
C1, C2 0, 0; 2.5 4, 3; 4

Notice player 2 does not have any information about nature or player 1 actions before the
game, so he should assign p=0.5 (i.e. total randomness).

Pure Bayesian-Nash equilibria: (A1, B2; D); (C1, C2; E)

b)

Looking at the normal form representations of the game, in game 1, B is strictly dominated by
A and, in game 2, E is strictly dominated by D and B dominates A.

(7,6)
D
2
E (2,0)
A
1
D (0,0)
C 2
E (4,4)
1-r
1 2 D
C (0,5)
r
B 2
D (10,8)
Microeconomics: Decision Theory, MQuEA Prof. Arguedas

Problem set #3
_____________________________________________________________________________

- Player 2, under Game 1, prefers (A,D) over (A,E), delivering payoffs (7,6).
- Player 2, under Game 2, always plays D if player 1 chooses B because it strictly dominates
E, delivering payoffs (10,8).
- If player 1 chooses C, player 2 goes for D if → 𝐸𝑈 (𝐷) ≥ 𝐸𝑈 (𝐸) ⇒ 0(1 − 𝑟) + 5𝑟 ≥

4 ⇒ 𝑟 ≥ 4 5. Therefore, if 𝑟 ≥ 4 5, player 1 gets 0 if he plays C, thus preferring (A,D)

under game 1 and (B,D) under game 2 because 7 and 10 are larger than 0. And if 𝑟 <
4 , player 1 gets 4 if he plays C, thus preferring (A,D) under game 1 and (B,D) under
5
game 2 because 7 and 10 are larger than 4.

With this information, player 2 can presume that player 1 will never choose C and that if player
1 goes for A, they are under Game 1, and if player 1 chooses B, they are under Game 2. Thus,
player 2 can use the action of player 1 as a signal for the Game type, delivering a separating
equilibrium.

The Perfect Bayesian Equilibrium is (A,D) for Game 1 and (B,D) for Game 2, a separating
equilibrium.

3. Consider the three following normal-form games, such that player 1 plays in rows and player
2 plays in columns:

GAME A
F G
F 1,1 4,0
G 0,4 3,3

GAME B
F G
F 3,1 0,0
G 0,0 1,3

GAME C
F G
F 1,0 0,1
G 0,1 1,0
Microeconomics: Decision Theory, MQuEA Prof. Arguedas

Problem set #3
_____________________________________________________________________________

(i) Assume that player 1 knows the payoff matrix of the game players are playing, but player 2
does not. In fact, player 2 only knows that he could be in game A or B or C with equal probability.
Assume also that player 1 plays first and player 2 can at least observe if player 1 has chosen
strategy F or G. Find the perfect Bayesian equilibrium of this game. Carefully explain the way in
which you find the equilibrium, and outline the equilibrium beliefs that sustain the equilibrium.
(ii) Assume that player 2 cannot observe player 1’s choice. Find the Bayesian Nash equilibrium
of this game. Carefully explain the way in which you find the equilibrium.

Solutions:

(i)

For player 1, strategy G is dominated by F in game 1.

(1,1)
F
2 G (4,0)

F 1/3 (3,1)
F
1 2 G
(0,0)
1/3 (1,0)
A
F F
2
1 G
(0,1)
B 1/3
G (0,0)
F
2
C F G
(1,3)
1/2
(0,1)
F
1 2
G G
(1,0)
1/2

With this information, player 2 is analysed first:

- If player 1 chooses F, player 2 goes for F if → 𝐸𝑈 (𝐹) ≥ 𝐸𝑈 (𝐺) ⇒ 1 3 1 + 1 3 1 +

1 0 ≥ 1 0 + 1 0 + 1 1 ⇒ 2 > 1 . Therefore, if player 1 chooses F, player 2
3 3 3 3 3 3
will go for F.
Microeconomics: Decision Theory, MQuEA Prof. Arguedas

Problem set #3
_____________________________________________________________________________

- If player 1 chooses G, player 2 goes for F if → 𝐸𝑈 (𝐹) ≥ 𝐸𝑈 (𝐺) ⇒ 1 2 0 + 1 2 1 ≥

1 3 + 1 0 ⇒ 1 < 3 . Therefore, if player 1 chooses G, player 2 will go for G.
2 2 2 2

Knowing the previous, under nature A, player 1 will choose (F,F). Under nature B, he will choose
(F,F) over (G,G) because payoffs are respectively (3,1) and (1,3). And under nature C, player 1
will be indifferent between (F,F) and (G,G) because payoffs are (1,0) in both cases.

The perfect Bayesian Equilibrium are (F,F) under Game A, (F,F) under Game B and both (F,F) and
(G,G) under Game 3.

(ii)

F G
FA, FB, FC 1, 3, 1; 2/3 1, 1, 0; 1/3
FA, FB, GC 1, 3, 0; 1 1, 1, 1; 0
FA, GB, GC 1, 0, 0; 2/3 1, 1, 1; 1
FA, GB, FC 1, 0, 1; 1/3 1, 1, 0; 4/3
GA, FB, FC 0, 3, 1; 5/3 3, 1, 0; 4/3
GA, GB, FC 0, 0, 1; 4/3 3, 1, 0; 7/3
GA, FB, GC 0, 3, 0; 2 3, 1, 1; 1
GA, GB, GC 0, 0, 0; 5/3 3, 1, 1; 2

As player 2 cannot observe what player 1 does, it is played as a simultaneous game. In order to
construct the table, first, we find all the combinations of strategies for player 1 and assign them
their payoffs.

For player 2, we compute the payoffs taking into account the probability of being either on A, B
or C game.

There are 2 Bayesian-Nash equilibrium: (FA, FB, FC; F) and (GA, GB, GC; G)

4. Two firms compete in quantities in a market of a product whose (inverse) aggregate demand
function is 𝑃 = 100 − 𝑄. Firm B knows that firm A operates with a marginal cost of 10. On the
other hand, firm A knows that the marginal cost of B is 10 on average, but firm B could be
Microeconomics: Decision Theory, MQuEA Prof. Arguedas

Problem set #3
_____________________________________________________________________________

operating with a high marginal cost of 16, or a low marginal cost of 4. Assume fixed costs are
zero for both firms. Compute the Bayesian Nash equilibrium of this Cournot game of incomplete
information.

Solution:

In the Cournot game, each firm 𝑖 has to maximize its profits by selecting the optimal quantity 𝑞 ∗
taken the others 𝑞 as given. We observe 2 types of B, so we will have to solve 2 problems for B
and 1 for A.
→ 𝐵 𝐻𝑖𝑔ℎ: max (100 − 𝑞 − 𝑞 )𝑞 −𝐶 ∙𝑞 ⇒ 𝑞 ∗ (𝑞 )

→ 𝐵 𝐿𝑜𝑤: max (100 − 𝑞 − 𝑞 )𝑞 −𝐶 ∙𝑞 ⇒ 𝑞 ∗ (𝑞 )

Firm A does not know which type is Firm B, so it has to compute its expected payoffs based on
the two possibilities of B, assigning the probabilities 𝛼 and (1 − 𝛼) to the two possible scenarios
𝐵 𝐻𝑖𝑔ℎ and 𝐵 𝐿𝑜𝑤:
𝛼∙𝐶 + (1 − 𝛼) ∙ 𝐶 =𝐶 , → 16𝛼 + 4(1 − 𝛼) = 10 → 𝛼 = 1 2
1 1
→ 𝐴: max ∙ [(100 − 𝑞 − 𝑞 )𝑞 − 𝐶 ∙ 𝑞 ] + ∙ [(100 − 𝑞 − 𝑞 )𝑞 − 𝐶 ∙ 𝑞 ]
2 2
The 3 maximizations provide us 3 equations for 3 variables, thus, the problem can be solved.
→ 𝐵 𝐻𝑖𝑔ℎ: max (100 − 𝑞 − 𝑞 )𝑞 − 16 ∙ 𝑞

𝜕𝜋 𝑞
= 0 = −(16 + 𝑞 ) − 2𝑞 ∗ + 100 ⇒ 𝑞 ∗ = 50 − 8 −
𝜕𝑞 2
→ 𝐵 𝐿𝑜𝑤: substitute 16 by 4, so 8 by 2: 𝑞 ∗ = 50 − 2 −
1 1
→ 𝐴: max ∙ [(100 − 𝑞 − 𝑞 )𝑞 − 10 ∙ 𝑞 ] + ∙ [(100 − 𝑞 − 𝑞 )𝑞 − 10 ∙ 𝑞 ]
2 2
1
= 100𝑞 − 𝑞 − 10𝑞 − 𝑞 (𝑞 + 𝑞 )
2
𝜕𝜋 1 1
= 0 ⇒ 100 = 2𝑞 ∗ + 10 + (𝑞 + 𝑞 ) ⇒ 𝑞 ∗ = 45 − (𝑞 + 𝑞 )
𝜕𝑞 2 4
Introducing 𝑞 ∗ and 𝑞 ∗ in 𝑞 ∗ :
1
𝑞 ∗ = 45 − (100 − 10 − 𝑞 ∗ ) ⇒ 𝑞 ∗ = 30
4
30
𝑞 ∗ = 50 − 8 − = 27
2
30
𝑞 ∗ = 50 − 2 − = 33
2
Bayesian-Nash Equilibria=(𝑞 ∗ , 𝑞 ∗ , 𝑞 ∗ ) = (30, 27, 33)
Microeconomics: Decision Theory, MQuEA Prof. Arguedas

Problem set #3
_____________________________________________________________________________

Furthermore, we can calculate prices for aggregated demand: (A, B Low)⇒ 𝑃 = 100 −
(30 + 33) = 37; (A, B High) ⇒ 𝑃 = 100 − (30 + 27) = 43

5. (This one resembles a poker game) The unique firm supplying a product in a market (firm A)
is threaten by the possible entrance of a competitor (firm B). If firm A operates with high costs,
its (monopolistic) profits are 4, but if firm B enters the market and competes against A, profits
for A and B are respectively 1 and 3. If instead firm A operates with low costs, its (monopolistic)
profits are 6, but if firm B enters the market and competes against A, profits for A and B are
respectively 3 and -1. Assume that B does not know whether firm A operates with high or low
costs. A way of signalling the cost structure to firm B is by setting a high or a low price. If A sets
a high price when costs are high and a low price when costs are low, profits for the two firms
are as expressed above. However, A can falsely set a low price when operating with high costs
to try to deter entry. In this case, firm A’s profits are 3 if it remains alone in the market, while all
this profit is captured by firm B under entry. Calculate the Perfect Bayesian equilibrium of this
game. Show that the equilibrium type (separating or pooling) crucially depends on the subjective
beliefs of firm B about the type of firm A.

Solution:

We can use a decision tree and backwards induction to solve this game:
(1,3)
E
B
NE (4,0)
PH
A
(0,3)
B E
PL
CH
1-q NE (3,0)
A
CL E (3,-1)
PL B

q (6,0)
NE
Microeconomics: Decision Theory, MQuEA Prof. Arguedas

Problem set #3
_____________________________________________________________________________

- If firm A chooses PL, firm B goes for E if → 𝐸𝑈 (𝐸) ≥ 𝐸𝑈 (𝑁𝐸) ⇒ 3(1 − 𝑞) − 1𝑞 ≥

0(1 − 𝑞) + 0𝑞 ⇒ 𝑞 ≤ 3 4. Therefore, if firm A chooses PL, firm B enters the market if

she thinks that firm A is falsely setting a low price with a probability larger than 25%.
- If firm A chooses PH, , firm B goes for E because she prefers payoffs (1,3) from (PH, E)
than payoffs (4,0) from (PH, NE).

After this analysis of firm B, if firm A manages to make 𝑞 > 3 4, under CH she will follow (PL, NE)
because payoffs (3,0) are higher for firm A than payoffs (1,3) from (PH, E). And under CL she will

follow (PL, NE). But if the best firm A can manage is to make 𝑞 ≤ 3 4, under CH she will follow
(PH, E) because payoffs (1,3) are higher for her than payoffs (0,3) from (PL, E). And under CL she
will follow (PL, E).

Hence, the Perfect Bayesian Equilibrium can be pooling or separating crucially depending on
firm B’s beliefs on the type of firm A:

- If 𝑞 > 3 4, the equilibrium is pooling: for CH, (PL, NE) and for CL, (PL, NE).

- If 𝑞 ≤ 3 4, the equilibrium is separating: for CH, (PH, E) and for CL, (PL, E).