National University of Singapore ECA5101 Microeconomic Analysis I

Department of Economics Semester 1 AY 2019/2020

Homework 4 — Solution
Dynamic Games and Moral Hazard

Due at the beginning of lecture 11 (week 12).

Question 1 (Sequential Split or Steal) Consider the “Split or Steal” game in homework 3.
Suppose Lucy moves first. Tony observes Lucy’s choice before making his decision. Draw
the extensive form of this game. Find all SPNE in this game.

The extensive form is shown above. Tony has two information sets and hence there are two
subgames. If Lucy chooses “Split”, then Tony should choose “Steal”. If Lucy chooses
“Steal”, Tony is indifferent between “Split” and “Steal”, therefore Tony can choose either.
Using backward induction, if Tony chooses “Steal if Split, Split if Steal”, Lucy will get 0 if
she chooses “Split” and 66885 if she chooses “Steal”, thus Lucy should choose “Steal. If
Tony chooses “Steal if Split, Steal if Steal”, Lucy will get 0 if she chooses “Split” and 0 if she
chooses “Steal”, thus Lucy is indifferent between “Split” and “Steal” and she can choose
either.

There are 3 pure strategy SPNE in this game:

Lucy chooses “Steal”, Tony chooses “Steal if Split, Split if Steal”.

Question 2 (The Chain Store Game) A chain store (player CS) has branches in K cities. In
each city k, there is a single potential competitor, player k. In period k, competitor k in city k
decides whether to compete with the chain store. The game goes like the following: in period
1, potential competitor in city 1 chooses whether to compete or not, “in” or “out”. If “in” is
chosen, the chain store then decides to “fight” or to “cooperate”. In period 2, the potential
competitor in city 2 observes the outcome in period 1 and chooses whether to compete or not,
and the chain store then respond to player 2’s decision, so on and so forth. The chain store
plays in all K periods, but each potential competitor only interacts with the chain store once.
(Hence this is not a repeated game since in each period there is a different potential
competitor.) The total payoff of the chain store is the sum of its payoffs in K cities. The
following figure summarizes the structure and the payoff in the stage game. The first number

!1
National University of Singapore ECA5101 Microeconomic Analysis I
Department of Economics Semester 1 AY 2019/2020

is the payoff to player k, the second number is the payoff to the chain store. Find all SPNE in

this game.

In the last period, period K, regardless of what has been played in the previous K-1 periods,
the chain store will choose to “cooperate if in” if player K chooses “in”, and knowing that
chain store will cooperate, player K will choose “in” rather than “out” because it gets 2
instead of 1. So the payoff to the chain store in period K is 2. In period K-1, the chain store
knows that in period K it will get a payoff of 2 no matter what the outcome is in period K-1.
Taking the payoff from period K into account, the reduced game in period K-1 is

The payoff for the chain store in the above game tree is the sum of payoff in the last 2
periods. Adding 2 to all possible payoffs of the chain store does not change the chain store’s
decision. If player K-1 chooses “in”, the chain store still chooses “cooperate if in” because its
payoff is 4, which is higher than 2. Knowing that the chain store will cooperate, player K-1
will choose “in”.

There is a unique SPNE in this game. In every period, the potential competitor chooses “in”
and the chain store chooses “cooperate if in”. The total payoff to the chain store is 2K.

Question 3 (Infinitely Repeated Cournot Game) Suppose the market demand for lumber is
given by P(Q) = 100 – Q/2. There are 2 symmetric producers in the market, each with the
following cost function TC(Q) = 10Q.

a) What is the joint monopoly price, quantity, and profits in this market?

!2
National University of Singapore ECA5101 Microeconomic Analysis I
Department of Economics Semester 1 AY 2019/2020

MR=100-Q=MC=10, thus Q=90, for each firm, Q=45. Monopoly price is P=55. Monopoly
profit is (55-10)*90=4050. Each firm’s profit is 2025.

b) Suppose in each period the two firms choose quantity simultaneously. If the game is
played only once, it is a standard Cournot competition. What is the Cournot equilibrium
quantity for each firm? What is the Cournot equilibrium profit for each firm?

Each firm’s reaction function is Q1=90-Q2/2. Thus the Cournot equilibrium quantity is
Q1=Q2=60. Cournot equilibrium price is P=40, and each firm earns profit (40-10)*60=1800.

c) Suppose the game is played infinitely many times. Each firm has a discount factor 0<d<1.
Suppose the two firms play the following Nash reversion strategy: each firm sets the joint
monopoly quantity as long as no one deviates, if either firm deviates by choosing some other
quantity, from next period on, both firms will set the static Cournot equilibrium quantity. At
some period T, firm 1 is contemplating on deviating. What quantity should firm 1 choose if it
deviates?

If firm 1 deviates, it should choose the best response to firm 2’s quantity (and firm 2 is not
deviating), 45. Thus firm 1 should choose Q1=90-Q2/2=90-45/2=67.5. If firm 1 chooses 67.5
and firm 2 chooses 45, market price is 43.75. Firm 1’s profit is (43.75-10)*67.5=2278.125.

d) For what values of d is each firm choosing the joint monopoly quantity in each period the
equilibrium outcome of an SPNE in this game?

At some period T, if firm 1 does not deviates, the sum of its discounted profit is
2025
!2025 + 2025d + 2025d 2 + . . . =
1−d

If firm 1 deviates, for that period, firm 1 gets a higher profit 2278.125, but from next period
on, firm 1’s profit will be 1800. The sum of its discounted profit is
1800d
!2278.125 + 1800d + 1800d 2 + . . . = 2278.125 +
1−d

To make sure firm 1 does not deviate, it must be that

which gives us d! ≥ 0.53 . Things are the same for firm 2. Thus if the discount factor is at
least 0.53, each firm choosing the joint monopoly quantity is the equilibrium outcome.

Question 4 (Cooperation in Finitely Repeated Game) Consider the following simultaneous-

!3
National University of Singapore ECA5101 Microeconomic Analysis I
Department of Economics Semester 1 AY 2019/2020

Player 2
A B C
A 10, 10 0, 20 0, 20
Player 1 B 20, 0 2, 2 1, 2
C 20, 0 2, 1 3, 3

a) Find all pure strategy Nash equilibrium in this game.

(B, B) and (C, C).

b) Suppose the game is repeated 20 times between the two players and the players do not
discount future payoffs. We want to know if it is possible to sustain the outcome (A, A) in
some periods of this game. Suppose each player adopts the following strategy. In period 1,
each player chooses A. If no one deviates, each player will choose C from period 2 onwards.
If someone deviates in period 1, that is, someone does not choose A in period 1, each player
will choose B from period 2 onwards. Using this strategy, is each player choosing A in period
1 the outcome of an SPNE?

In any period after period 1, if (A, A) is played in period 1, according to the strategy, (C, C)
will be played and this is a Nash equilibrium for the subgame because (C, C) is a Nash
equilibrium in the stage game. Similarly, if (A, A) is not played in period 1, (B, B) will be
played and this is a Nash equilibrium for the subgame.

In period 1, if player 1 chooses A, the total payoff is 10+3*19=67. If player 1 deviates, he

Since there are more than one Nash equilibria in the stage game, punishment can be credible
even in a finitely repeated game — if we cooperate in period 1, we will play the good Nash
equilibrium in the remaining periods, if someone deviates in period 1, we will play the bad
Nash equilibrium in the remaining periods. Playing the bad Nash equilibrium in the
remaining periods is the punishment for cheating. It is credible because it is a stage game
Nash equilibrium.

Question 5 (Principal-Agent Model with Three Effort Levels) A plaintiff hires a lawyer for a
case. The lawyer can choose low effort, medium effort, or high effort. If the lawyer chooses
low effort, the probability of winning the case is 0.5. If the lawyer chooses medium effort, the
probability of winning the case is 0.6. If the lawyer chooses high effort, the probability of
winning the case is 0.8. Low effort costs the lawyer $500, medium effort costs the lawyer
$700, and high effort costs the lawyer $1000. If the plaintiff wins the case, he will receive
$10,000 from the defendant. If the plaintiff loses the case, he gets 0 from the defendant (and
he also does not need to pay anything to the defendant). Suppose both the plaintiff and the
lawyer are risk neutral. The lawyer’s reservation utility is 0.

!4
National University of Singapore ECA5101 Microeconomic Analysis I
Department of Economics Semester 1 AY 2019/2020

a) What is the first-best effort level? What is the first-best payoff for the plaintiff?

If the plaintiff wants low effort, he needs to pay the lawyer $500 for low effort and 0
otherwise. The lawyer will choose low effort and the plaintiff’s expected payoff is
0.5*10,000-500=4500.

If the plaintiff wants medium effort, he has to pay the lawyer $700 for medium effort and 0
otherwise. The lawyer will choose medium effort and the plaintiff’s expected payoff is
0.6*10,000-700=5300.

If the plaintiff wants high effort, he should pay $1000 to the lawyer for high effort and 0
otherwise. The lawyer will choose high effort, and the plaintiff’s expected payoff is
0.8*10,000-1000=7000.

Thus the first-best effort level is high effort. The expected payoff for the plaintiff is $7000.

b) Suppose the plaintiff can sell the case to the lawyer for a fixed fee. How much should the
fixed fee be? What is the expected payoff for the plaintiff?

Let the fee be a. If the lawyer chooses low effort, his expected utility is 0.5*10000-
a-500=4500-a. If the lawyer chooses medium effort, his expected utility is 0.6*10000-
a-700=5300-a. If the lawyer chooses high effort, his expected utility is 0.8*10000-
a-1000=7000-a. Thus the lawyer will choose high effort. The principal should set a=7000.
The expected payoff for the principal is $7000.

c) Suppose selling the case to the lawyer is prohibited by law. Consider the contract where
the plaintiff pays the lawyer only when he wins the case. For the lawyer to choose the first-
best effort, how much should the plaintiff pay the lawyer? What is the expected payoff for the
plaintiff?

Let the wage for the lawyer be x. If the lawyer chooses high effort, his expected utility is
0.8x-1000. If the lawyer chooses low effort, his expected utility is 0.5x-500. If the lawyer
chooses medium effort, his expected utility is 0.6x-700. For the lawyer to choose high effort,
the IR is 0.8x-1000>=0, and the ICs are 0.8x-1000>=0.5x-500, and 0.8x-1000>=0.6x-700.
Solving the IR and the ICs we get x>=1667. The plaintiff maximizes 0.8(10000-x). Thus the
plaintiff will pay $1667 to the lawyer in the event of winning and the lawyer will choose high
effort. The plaintiff’s expected payoff is 0.8(10000-1667)=$6666.

!5