Problem set 1

1. Consider the infinitely repeated game with discount rate δ, where the strategic form below
is the stage game. Show that for sufficiently patient players, it is a subgame perfect Nash
equilibrium to play (U, L) in every period. What is the minimum δ that achieves cooperation?

B
L R
A U 1,1 -2,5
D 2,0 0,0

Solution:

The Nash equilibrium of the stage game is (D, R), giving payoffs (0, 0). Consider the trigger
strategies.

If the history is (U, L),(U, L), ...,(U, L), play (U, L) this period.
After any other history, play (D, R)

Does either player have a profitable deviation? For the row player, following the proposed
strategies gives a payoff of

1
1 + 𝛿. 1 + 𝛿 2 . 1 + ⋯ =
1−𝛿
and deviating gives a payoff of

2 + 𝛿0 + 𝛿 2 0 + ⋯ = 2

so cooperating is better than deviating if

1
≥2
1−𝛿

Or

1
𝛿≥
2

For the column player, following the proposed strategies gives a payoff of

1
1 + 𝛿. 1 + 𝛿 2 . 1 + ⋯ =
1−𝛿
and deviating gives a payoff of

5 + 𝛿. 0 + 𝛿 2 . 0 + ⋯ = 5

So cooperating is better than deviating if

 1
≥5
1−𝛿
Or
4
𝛿≥
5
Therefore, as long as δ ≥ 4/5, the trigger strategies are a Subgame Perfect Nash Equilibrium of the
infinitely repeated game.

2. Consider an infinitely repeated Cournot duopoly game with linear demand function, 𝑄 =
𝛼 − 𝑃 and constant marginal cost, c >0. Is it possible to sustain the monopoly outcome
(each producing half of the monopoly quantity) in SPE?

Solution:

Let us consider the following inverse demand function.

𝑃 = 𝛼 − 𝑄, where, 𝑄 = 𝑞𝑖 + 𝑞−𝑖
(𝛼−𝑐)2
Payoffs under Cournot duopoly setup is 9

(𝛼−𝑐)2
Payoffs under Monopoly outcomes is 8

B
Collude Not collude
A Collude (𝛼 − 𝑐)2 (𝛼 − 𝑐)2 3(𝛼 − 𝑐)2 9(𝛼 − 𝑐)2
, ,
8 8 32 64

Not collude 9(𝛼 − 𝑐)2 3(𝛼 − 𝑐)2 (𝛼 − 𝑐)2 (𝛼 − 𝑐)2

, ,
64 32 9 9

The Nash equilibrium of the stage game is (Not collude, Not collude), giving payoffs
(𝛼−𝑐)2 (𝛼−𝑐)2
( 9
, 9
). Consider the trigger strategies.

If the history is (Collude, Collude), (Collude, Collude), ..., (Collude, Collude), play (Collude,
Collude) this period.
After any other history, play (Not collude, Not collude)

For the any firm, following the proposed strategies gives a payoff of

(𝛼 − 𝑐)2
(𝛼 − 𝑐)2 (𝛼 − 𝑐)2 (𝛼 − 𝑐)2 8
+ 𝛿. + 𝛿2. +⋯=
8 8 8 1−𝛿

and deviating gives a payoff of

 (𝛼 − 𝑐)2 𝛿
9(𝛼 − 𝑐)2 (𝛼 − 𝑐)2 (𝛼 − 𝑐)2 9(𝛼 − 𝑐)2 9
+𝛿 + 𝛿2 +⋯ = +
64 9 9 64 1−𝛿

so cooperating is better than deviating if

1 9 𝛿
≥ +
8(1 − 𝛿) 64 9(1 − 𝛿)

9
Therefore, ≥ 17 , the trigger strategies are a Subgame Perfect Nash Equilibrium of the infinitely
repeated game.

3. Consider Bertrand’s model of duopoly in the case that each firm’s unit cost is constant, equal
to c. Let 𝜋 𝑚 (𝑝) = (𝑝 − 𝑐)𝐷(𝑝) for any price p,and assume that Π is continuous and is
uniquely maximized at the price p m (the “monopoly price”). Let s be the strategy for the
infinitely repeated game that charges 𝑝𝑚 in the first period and subsequently as long as
both firms continue to charge p m, and punishes any deviation from 𝑝𝑚 by either firm by
choosing the price c for k periods, then reverting to 𝑝𝑚 . Given any value of δ, for what
values of k is the strategy pair (s, s) a subgame perfect equilibrium of the infinitely repeated
game?

Solution:

Suppose that firm i uses the strategy 𝑠𝑖 . If the other firm, j, uses 𝑠𝑗 , then its discounted
average payoff is
1 1 1
(1 − 𝛿) ( 𝜋 𝑚 (𝑝) + 𝛿𝜋 𝑚 (𝑝) + ⋯ ) = 𝜋 𝑚 (𝑝)
2 2 2

If on the other hand, firm j deviates to a price p close to 𝑝𝑚 and obtain profits close to
𝜋 𝑚 in the period of deviation. It’s profit during k periods of punishment is zero. Once
the punishment is over, it can either go back to 𝑝𝑚 or deviate.

Therefore, the maximum profit deviating is as follows

(1 − 𝛿)𝜋 𝑚 (𝑝)
(1 − 𝛿)( 𝜋 𝑚 (𝑝) + 𝛿 𝑘+1 𝜋 𝑚 (𝑝) + 𝛿 2𝑘+2 𝜋 𝑚 (𝑝) … ) =
(1 − 𝛿𝑘+1 )

(1 − 𝛿) 1
𝑘+1
≤
(1 − 𝛿 ) 2

𝛿 𝑘+1 − 2𝛿 + 1 ≤ 0
 4. Consider the infinitely repeated Prisoner’s Dilemma in which the payoffs of the component
game are those given in the figure below.

2
C D
1 C x,x 0,x+1
D x+1,0 y,y

Consider the strategy following tit-for-tat strategy:

𝐶 if 𝑡 = 0 or ℎ𝑗𝑡 = 𝐶
𝑠𝑖 (ℎ𝑡 ) =
𝐷 if 𝑡 > 0 and ℎ𝑗𝑡 = 𝐷

Under which conditions on the parameters x > y > 0, and 0 < δ < 1 is the strategy profile (tit-
for-tat, tit-for-tat) a subgame perfect equilibrium?

Solution

If history at t is (C,D), then the sequence of play from t+1 is (D,C) (D,C), (D,C) (D,C),…

(𝑥+1) 𝛿.(𝑥+1)
and payoffs are ((1−𝛿 2 ) , (1−𝛿 2 ) ).

If history at t is (D,C), then the sequence of play from t+1 is (D,C), (D,C) (D,C), (D,C), …

𝛿.(𝑥+1) (𝑥+1)
and payoffs are ( (1−𝛿 2 ) , (1−𝛿 2 )).

𝑦 𝑦
After (D,D) at t, (D,D) will be played thereafter, resulting in ((1−𝛿) , (1−𝛿)).

Given the above continuation games, the reduced game at any  for any previous history is

2
C D
1 C 𝑥𝛿 𝑥𝛿 (𝑥 + 1)𝛿 (𝑥 + 1)𝛿
𝑥+ ,𝑥 + 0+ 2
, (𝑥 + 1) + 𝛿
(1 − 𝛿) (1 − 𝛿) (1 − 𝛿 ) (1 − 𝛿 2 )
D (𝑥 + 1)𝛿 (𝑥 + 1)𝛿 𝑦𝛿 𝑦𝛿
(𝑥 + 1) + 𝛿 2
,0+ 𝑦+ ,𝑦 +
(1 − 𝛿 ) (1 − 𝛿 2 ) (1 − 𝛿) (1 − 𝛿)

Consider t = 0, when (Tit-for-tat,Tit-for-tat) prescribes (C,C). Single-deviation principle then

requires that (C,C) is a Nash equilibrium of the reduced game above. That is, we must have

𝑥 + 𝛿𝑥 + 𝛿 2 𝑥 + ⋯ ≥ (𝑥 + 1) + 𝛿. 0 + 𝛿 2 (𝑥 + 1) + 𝛿 3 . 0 + ⋯
 𝑥𝛿 (𝑥 + 1)𝛿
𝑥+ ≥ (𝑥 + 1) + 𝛿
(1 − 𝛿) (1 − 𝛿 2 )

𝑥 (𝑥+1) 𝛿.0
Since , (1−𝛿) ≥ (1−𝛿 2 ) + (1−𝛿 2 )

𝛿 ≥ 1/𝑥

Consider a history in which (C,D) is played at t − 1. Now according to (Tit-fortat,Tit-for-tat)

we must have (D,C) at t. Single-deviation principle now requires that (D,C) is a Nash
equilibrium of the above game. That is, we must have

𝑥 (𝑥 + 1) 𝛿. 0
≤ +
(1 − 𝛿) (1 − 𝛿 ) (1 − 𝛿 2 )
2

the opposite of the previous requirement. Hence, (Tit-for-tat, Tit-for-tat) is not a subgame-
perfect Nash equilibrium, unless

𝛿 = 1/𝑥

Likewise,

Consider a history in which (D,D) is played at t − 1. Now according to (Tit-for-tat, Tit-for-tat)

we must have (D,D) at t. Single-deviation principle now requires that (D,D) is a Nash
equilibrium of the above game. That is, we must have

𝑦 + 𝛿𝑦 + 𝛿 2 𝑦 + ⋯ ≥ 0 + 𝛿. (1 + 𝑥) + 𝛿 2 0 + 𝛿 3 . (𝑥 + 1) + ⋯

𝑦 0 𝛿. (𝑥 + 1)
≥ 2
+
(1 − 𝛿) (1 − 𝛿 ) (1 − 𝛿 2 )

𝑦
𝛿≤
𝑥−𝑦+1

Consider a history in which (D,C) is played at t − 1. Now according to (Tit-for-tat, Tit-for-tat)

we must have (C,D) at t. Single-deviation principle now requires that (C,D) is a Nash
equilibrium of the above game. That is, we must have

0 + 𝛿. (1 + 𝑥) + 𝛿 2 0 + 𝛿 3 . (𝑥 + 1) + ⋯ ≥ 𝑦 + 𝛿𝑦 + 𝛿 2 𝑦 + ⋯
 𝑦
𝛿≥
𝑥−𝑦+1

Therefore only when,

𝑦
𝛿= = 1/𝑥 , tit-for-tat is SPNE.
𝑥−𝑦+1