Lecture 18: Repeated Games

Mauricio Romero
Lecture 18: Repeated Games

Recap from last class

More than one NE in the stage game

Example 1

Example 2
Lecture 18: Repeated Games

Recap from last class

More than one NE in the stage game

Example 1

Example 2
Theorem
Suppose that the stage game G has exactly one NE, (a1∗ , a2∗ , . . . , an∗ ). Then for any
δ ∈ (0, 1] and any T , the T -times repeated game has a unique SPNE in which all
players i play ai∗ at all information sets.
I The basic idea of the proof for this proposition is exactly the same that we saw in
the repeated prisoner’s dilemma
I The basic idea of the proof for this proposition is exactly the same that we saw in
the repeated prisoner’s dilemma
I All past payoffs are sunk
I The basic idea of the proof for this proposition is exactly the same that we saw in
the repeated prisoner’s dilemma
I All past payoffs are sunk
I In the last period, the incentives of all players are exactly the same as if the game
were being played once
I The basic idea of the proof for this proposition is exactly the same that we saw in
the repeated prisoner’s dilemma
I All past payoffs are sunk
I In the last period, the incentives of all players are exactly the same as if the game
were being played once
I Thus all players must play the stage game Nash equilibrium action regardless of
the history of play up to that point
I The basic idea of the proof for this proposition is exactly the same that we saw in
the repeated prisoner’s dilemma
I All past payoffs are sunk
I In the last period, the incentives of all players are exactly the same as if the game
were being played once
I Thus all players must play the stage game Nash equilibrium action regardless of
the history of play up to that point
I But then we can induct
I The basic idea of the proof for this proposition is exactly the same that we saw in
the repeated prisoner’s dilemma
I All past payoffs are sunk
I In the last period, the incentives of all players are exactly the same as if the game
were being played once
I Thus all players must play the stage game Nash equilibrium action regardless of
the history of play up to that point
I But then we can induct
I Knowing that the stage game Nash equilibrium is going to be played tomorrow, at
any information set, we can ignore the past payoffs
I The basic idea of the proof for this proposition is exactly the same that we saw in
the repeated prisoner’s dilemma
I All past payoffs are sunk
I In the last period, the incentives of all players are exactly the same as if the game
were being played once
I Thus all players must play the stage game Nash equilibrium action regardless of
the history of play up to that point
I But then we can induct
I Knowing that the stage game Nash equilibrium is going to be played tomorrow, at
any information set, we can ignore the past payoffs
I We concentrate just on the payoffs in the future. Thus in period T − 1, player i
simply wants to maximize:
T −1
max δ T −2 ui (ai , a−i ) + δ T −1 ui (a∗ ).
ai ∈Ai
I What player i plays today has no consequences for what happens in period T
since we saw that all players will play a∗ no matter what happens in period T − 1
I What player i plays today has no consequences for what happens in period T
since we saw that all players will play a∗ no matter what happens in period T − 1

I So, the maximization problem above is the same as:

T −1
max ui (ai , a−i ).
ai ∈Ai
I What player i plays today has no consequences for what happens in period T
since we saw that all players will play a∗ no matter what happens in period T − 1

I So, the maximization problem above is the same as:

T −1
max ui (ai , a−i ).
ai ∈Ai

I Thus again, for this to be a Nash equilibrium, we need a1T −1 = a1∗ , . . . , anT −1 = an∗ .
I What player i plays today has no consequences for what happens in period T
since we saw that all players will play a∗ no matter what happens in period T − 1

I So, the maximization problem above is the same as:

T −1
max ui (ai , a−i ).
ai ∈Ai

I Thus again, for this to be a Nash equilibrium, we need a1T −1 = a1∗ , . . . , anT −1 = an∗ .

I Following exactly this induction, we can conclude that every player must play ai∗
at all times and all histories
Lecture 18: Repeated Games

Recap from last class

More than one NE in the stage game

Example 1

Example 2
Lecture 18: Repeated Games

Recap from last class

More than one NE in the stage game

Example 1

Example 2
I What would happen if there are more than one NE of the stage game?
I What would happen if there are more than one NE of the stage game?

I Suppose instead that the stage game looks as follows

Normal Form

A2 B2 C2
A1 1, 1 0, 0 0, 0
B1 0, 0 4, 4 1, 5
C1 0, 0 5, 1 3, 3
I If the game is only played once
I If the game is only played once

I There are two pure strategy Nash equilibria: (A1 , A2 ) and (C1 , C2 ).
I If the game is only played once

I There are two pure strategy Nash equilibria: (A1 , A2 ) and (C1 , C2 ).

I (B1 , B2 ) is not a Nash equilibrium if the game is only played once

I If the game is only played once

I There are two pure strategy Nash equilibria: (A1 , A2 ) and (C1 , C2 ).

I (B1 , B2 ) is not a Nash equilibrium if the game is only played once

I In the one-shot game, the Nash equilibria are inefficient because they are Pareto
dominated by (B1 , B2 )
I Playing the NE of the stage game in every period is a SPNE in the repeated game
I Playing the NE of the stage game in every period is a SPNE in the repeated game

I The logic is the same as when there is a single NE

I Always playing (A1 , A2 ) is a SPNE
I Always playing (A1 , A2 ) is a SPNE

I Player 1’s strategy is given by:

1. Play A1 in period 1;
2. Play A1 at all histories in period 2.

I Player 2’s strategy is given by:

1. Play A2 in period 1;
2. Play A2 at all histories in period 2.
I Always playing (C1 , C2 ) is a SPNE
I Always playing (C1 , C2 ) is a SPNE

I Player 1’s strategy is given by:

1. Play C1 in period 1;
2. Play C1 at all histories in period 2.

I Player 2’s strategy is given by:

1. Play C2 in period 1;
2. Play C2 at all histories in period 2.
But are there more?
I Combining NE of the stage game is also a SPNE
I Combining NE of the stage game is also a SPNE

I The logic is the same as before

I Playing (A1 , A2 ) in t = 1 and (C1 , C2 ) in t = 2 is a SPNE
I Playing (A1 , A2 ) in t = 1 and (C1 , C2 ) in t = 2 is a SPNE

I Player 1’s strategy is given by:

1. Play A1 in period 1;
2. Play C1 at all histories in period 2.

I Player 2’s strategy is given by:

1. Play A2 in period 1;
2. Play C1 at all histories in period 2.
I Similarly, playing (C1 , C2 ) in t = 1 and (A1 , A2 ) in t = 2 is a SPNE
I Similarly, playing (C1 , C2 ) in t = 1 and (A1 , A2 ) in t = 2 is a SPNE

I Player 1’s strategy is given by:

1. Play C1 in period 1;
2. Play A1 at all histories in period 2.

I Player 2’s strategy is given by:

1. Play C2 in period 1;
2. Play A1 at all histories in period 2.
I This is uninteresting since Nash equilibria are played in every period
I This is uninteresting since Nash equilibria are played in every period

I But are there more?

I This is uninteresting since Nash equilibria are played in every period

I But are there more?

I The SPNE that we’ve considered, players always play strategies that do not
condition on what happened in the past
I This is uninteresting since Nash equilibria are played in every period

I But are there more?

I The SPNE that we’ve considered, players always play strategies that do not
condition on what happened in the past

I What makes a repeated game interesting is when players play strategies in SPNE
that condition on what happened in the past
I This is uninteresting since Nash equilibria are played in every period

I But are there more?

I The SPNE that we’ve considered, players always play strategies that do not
condition on what happened in the past

I What makes a repeated game interesting is when players play strategies in SPNE
that condition on what happened in the past

I This could not happen when the stage game had a unique NE
I This is uninteresting since Nash equilibria are played in every period

I But are there more?

I The SPNE that we’ve considered, players always play strategies that do not
condition on what happened in the past

I What makes a repeated game interesting is when players play strategies in SPNE
that condition on what happened in the past

I This could not happen when the stage game had a unique NE

I In the last period, all players were required to play the unique NE action after all
histories!
I This is uninteresting since Nash equilibria are played in every period

I But are there more?

I The SPNE that we’ve considered, players always play strategies that do not
condition on what happened in the past

I What makes a repeated game interesting is when players play strategies in SPNE
that condition on what happened in the past

I This could not happen when the stage game had a unique NE

I In the last period, all players were required to play the unique NE action after all
histories!
I This is uninteresting since Nash equilibria are played in every period

I But are there more?

I The SPNE that we’ve considered, players always play strategies that do not
condition on what happened in the past

I What makes a repeated game interesting is when players play strategies in SPNE
that condition on what happened in the past

I This could not happen when the stage game had a unique NE

I In the last period, all players were required to play the unique NE action after all
histories! Why?
Proof

I To see this, suppose that a history (a1 , a2 ) was played in period 1 resulting in
payoffs from period 1 of (x, y )
Proof

I To see this, suppose that a history (a1 , a2 ) was played in period 1 resulting in
payoffs from period 1 of (x, y )

I Then the normal form of the subgame starting in period 2 is given by:
Normal Form

A2 B2 C2
A1 (x, y ) + δ(1, 1) (x, y ) + δ(0, 0) (x, y ) + δ(0, 0)
B1 (x, y ) + δ(0, 0) (x, y ) + δ(4, 4) (x, y ) + δ(1, 5)
C1 (x, y ) + δ(0, 0) (x, y ) + δ(5, 1) (x, y ) + δ(3, 3)
Proof

I Since we are just adding the same (x, y ) to each cell and multiplying by δ, the
Nash equilibrium remains unchanged from the original stage game
Proof

I Since we are just adding the same (x, y ) to each cell and multiplying by δ, the
Nash equilibrium remains unchanged from the original stage game

I The set of Nash equilibria of this subgame is given by (A1 , A2 ) and (C1 , C2 )
Proof

I Since we are just adding the same (x, y ) to each cell and multiplying by δ, the
Nash equilibrium remains unchanged from the original stage game

I The set of Nash equilibria of this subgame is given by (A1 , A2 ) and (C1 , C2 )

I Thus after any history, the set of pure strategy NE are (A1 , A2 ) or (C1 , C2 )
Proof

I Since we are just adding the same (x, y ) to each cell and multiplying by δ, the
Nash equilibrium remains unchanged from the original stage game

I The set of Nash equilibria of this subgame is given by (A1 , A2 ) and (C1 , C2 )

I Thus after any history, the set of pure strategy NE are (A1 , A2 ) or (C1 , C2 )

I Since SPNE requires Nash equilibrium in every subgame, this means that after
any history, (A1 , A2 ) or (C1 , C2 ) must be played
I Lets try to find a SPNE in which (B1 , B2 ) is played in the first period.

Normal Form

A2 B2 C2
A1 1, 1 0, 0 0, 0
B1 0, 0 4, 4 1, 5
C1 0, 0 5, 1 3, 3
I Consider the following strategy profile, where we punish in t = 2 if we don’t play
(B1 , B2 ) in t = 1
I Consider the following strategy profile, where we punish in t = 2 if we don’t play
(B1 , B2 ) in t = 1

I Anna plays the following strategy:

I Consider the following strategy profile, where we punish in t = 2 if we don’t play
(B1 , B2 ) in t = 1

I Anna plays the following strategy:

1. Play B1 in period 1.
I Consider the following strategy profile, where we punish in t = 2 if we don’t play
(B1 , B2 ) in t = 1

I Anna plays the following strategy:

1. Play B1 in period 1.
2. Play A1 in period 2 if anything other than (B1 , B2 ) is played in period 1,
I Consider the following strategy profile, where we punish in t = 2 if we don’t play
(B1 , B2 ) in t = 1

I Anna plays the following strategy:

1. Play B1 in period 1.
2. Play A1 in period 2 if anything other than (B1 , B2 ) is played in period 1,
3. Play C1 in period 2 if (B1 , B2 ) is played in period 1.
I Consider the following strategy profile, where we punish in t = 2 if we don’t play
(B1 , B2 ) in t = 1

I Anna plays the following strategy:

1. Play B1 in period 1.
2. Play A1 in period 2 if anything other than (B1 , B2 ) is played in period 1,
3. Play C1 in period 2 if (B1 , B2 ) is played in period 1.

I Bob plays a similar strategy:

I Consider the following strategy profile, where we punish in t = 2 if we don’t play
(B1 , B2 ) in t = 1

I Anna plays the following strategy:

1. Play B1 in period 1.
2. Play A1 in period 2 if anything other than (B1 , B2 ) is played in period 1,
3. Play C1 in period 2 if (B1 , B2 ) is played in period 1.

I Bob plays a similar strategy:

1. Play B2 in period 1.
I Consider the following strategy profile, where we punish in t = 2 if we don’t play
(B1 , B2 ) in t = 1

I Anna plays the following strategy:

1. Play B1 in period 1.
2. Play A1 in period 2 if anything other than (B1 , B2 ) is played in period 1,
3. Play C1 in period 2 if (B1 , B2 ) is played in period 1.

I Bob plays a similar strategy:

1. Play B2 in period 1.
2. Play A2 in period 2 if anything other than (B1 , B2 ) is played in period 1,
I Consider the following strategy profile, where we punish in t = 2 if we don’t play
(B1 , B2 ) in t = 1

I Anna plays the following strategy:

1. Play B1 in period 1.
2. Play A1 in period 2 if anything other than (B1 , B2 ) is played in period 1,
3. Play C1 in period 2 if (B1 , B2 ) is played in period 1.

I Bob plays a similar strategy:

1. Play B2 in period 1.
2. Play A2 in period 2 if anything other than (B1 , B2 ) is played in period 1,
3. Play C2 in period 2 if (B1 , B2 ) is played in period 1.
If (B1 , B2 ) is observed in the first period, the subgame corresponding to that
observation admits the following normal form:

Normal Form

A2 B2 C2
A1 (4, 4) + δ(1, 1) (4, 4) + δ(0, 0) (4, 4) + δ(0, 0)
B1 (4, 4) + δ(0, 0) (4, 4) + δ(4, 4) (4, 4) + δ(1, 5)
C1 (4, 4) + δ(0, 0) (4, 4) + δ(5, 1) (4, 4) + δ(3, 3)
I The subgame is just the original game with a payoff of (4, 4) added to each box
and multiplying by δ
I The subgame is just the original game with a payoff of (4, 4) added to each box
and multiplying by δ

I If we add the same utility to all boxes, then the preferences of players are
completely unchanged
I The subgame is just the original game with a payoff of (4, 4) added to each box
and multiplying by δ

I If we add the same utility to all boxes, then the preferences of players are
completely unchanged

I Therefore the set of Nash equilibria are the same in this subgame as in the stage
game
I The subgame is just the original game with a payoff of (4, 4) added to each box
and multiplying by δ

I If we add the same utility to all boxes, then the preferences of players are
completely unchanged

I Therefore the set of Nash equilibria are the same in this subgame as in the stage
game

I So it is a Nash equilibrium in this subgame for players to play (A1 , A2 ), which is

consistent with the strategy that we proposed
I Let us now check that after observing (α1 , α2 ) 6= (B1 , B2 ), then it is a Nash
equilibrium in the subgame for players to play (C1 , C2 )
I Let us now check that after observing (α1 , α2 ) 6= (B1 , B2 ), then it is a Nash
equilibrium in the subgame for players to play (C1 , C2 )

I If (α1 , α2 ) 6= (B1 , B2 ) is observed there are some payoffs (x, y ) such that the
subgame induces the following normal form
Normal Form

A2 B2 C2
A1 (x, y ) + δ(1, 1) (x, y ) + δ(0, 0) (x, y ) + δ(0, 0)
B1 (x, y ) + δ(0, 0) (x, y ) + δ(4, 4) (x, y ) + δ(1, 5)
C1 (x, y ) + δ(0, 0) (x, y ) + δ(5, 1) (x, y ) + δ(3, 3)
I Again in this case, note that we are simply adding the same payoff profile (x, y ) to
every box and multiplying by δ
I Again in this case, note that we are simply adding the same payoff profile (x, y ) to
every box and multiplying by δ

I Therefore, the Nash equilibrium is again the set of Nash equilibrium of the original
stage game
I Again in this case, note that we are simply adding the same payoff profile (x, y ) to
every box and multiplying by δ

I Therefore, the Nash equilibrium is again the set of Nash equilibrium of the original
stage game

I In this subgame, it is a Nash equilibrium for players to play (A1 , A2 )

I We have checked that the strategy profile was indeed a Nash equilibrium in all
subgames that begin in period 2
I We have checked that the strategy profile was indeed a Nash equilibrium in all
subgames that begin in period 2

I The only other subgame is the whole game itself

I We have checked that the strategy profile was indeed a Nash equilibrium in all
subgames that begin in period 2

I The only other subgame is the whole game itself

I We need to check that indeed the strategies constitute a Nash equilibrium in the
whole game
I We have checked that the strategy profile was indeed a Nash equilibrium in all
subgames that begin in period 2

I The only other subgame is the whole game itself

I We need to check that indeed the strategies constitute a Nash equilibrium in the
whole game

I To do this, we already specified the play at all information sets in the second
period
So we can simplify the game which gives the following game tree.

Anna

A1 C1
B1
Bob Bob Bob

A2 C2 A2 C2
B2 B2 A2 C2
B2
(1 + δ, 1 + δ) (δ, δ) (δ, δ) (1 + δ, 5 + δ)

(δ, δ) (4 + 3δ, 4 + 3δ) (δ, δ) (3 + δ, 3 + δ)

(5 + δ, 1 + δ)
The normal form of this game (conditional on what happens in T = 2) is:

Normal Form

A2 B2 C2
A1 1 + δ, 1 + δ δ, δ δ, δ
B1 δ, δ 4 + 3δ, 4 + 3δ 1 + δ, 5 + δ
C1 δ, δ 5 + δ, 1 + δ 3 + δ, 3 + δ
I In this game the best response for player i is:



Ai if s−i = A−i

B
i if s−i = B−i & 4 + 3δ ≥ 5 + δ
BRi (s−i ) =


Ci if s−i = B−i & 4 + 3δ ≤ 5 + δ

C
i if s−i = C−i
I In this game the best response for player i is:



Ai if s−i = A−i

B
i if s−i = B−i & 4 + 3δ ≥ 5 + δ
BRi (s−i ) =


Ci if s−i = B−i & 4 + 3δ ≤ 5 + δ

C
i if s−i = C−i

I (B1 , B2 ) is a Nash equilibrium if 4 + 3δ ≥ 5 + δ

I In this game the best response for player i is:



Ai if s−i = A−i

B
i if s−i = B−i & 4 + 3δ ≥ 5 + δ
BRi (s−i ) =


Ci if s−i = B−i & 4 + 3δ ≤ 5 + δ

C
i if s−i = C−i

I (B1 , B2 ) is a Nash equilibrium if 4 + 3δ ≥ 5 + δ

I (B1 , B2 ) is a Nash equilibrium if δ > 1/2
I In this game the best response for player i is:



Ai if s−i = A−i

B
i if s−i = B−i & 4 + 3δ ≥ 5 + δ
BRi (s−i ) =


Ci if s−i = B−i & 4 + 3δ ≤ 5 + δ

C
i if s−i = C−i

I (B1 , B2 ) is a Nash equilibrium if 4 + 3δ ≥ 5 + δ

I (B1 , B2 ) is a Nash equilibrium if δ > 1/2
I The strategy profile defined for Anna and Bob at the beginning of this section is
indeed a subgame perfect Nash equilibrium if players value the future enough
(δ > 1/2)
I In this game the best response for player i is:



Ai if s−i = A−i

B
i if s−i = B−i & 4 + 3δ ≥ 5 + δ
BRi (s−i ) =


Ci if s−i = B−i & 4 + 3δ ≤ 5 + δ

C
i if s−i = C−i

I (B1 , B2 ) is a Nash equilibrium if 4 + 3δ ≥ 5 + δ

I (B1 , B2 ) is a Nash equilibrium if δ > 1/2
I The strategy profile defined for Anna and Bob at the beginning of this section is
indeed a subgame perfect Nash equilibrium if players value the future enough
(δ > 1/2)
I If players value the future enough (δ > 1/2), then the future prize is worth the
short term loss
I What is the take away of this exercise?
I What is the take away of this exercise?
I In the repeated Prisoner’s Dilemma, the stage game (played just once) had just
one Nash equilibrium
I What is the take away of this exercise?
I In the repeated Prisoner’s Dilemma, the stage game (played just once) had just
one Nash equilibrium
I The only subgame perfect Nash equilibrium was to play the Nash equilibrium of
the stage game in every period
I What is the take away of this exercise?
I In the repeated Prisoner’s Dilemma, the stage game (played just once) had just
one Nash equilibrium
I The only subgame perfect Nash equilibrium was to play the Nash equilibrium of
the stage game in every period
I In fact, one can prove generally that if the stage game has only one Nash
equilibrium then in the repeated game with that stage game, the unique subgame
perfect Nash equilibrium requires the Nash equilibrium to be played in all periods
and all information sets
I What is the take away of this exercise?
I In the repeated Prisoner’s Dilemma, the stage game (played just once) had just
one Nash equilibrium
I The only subgame perfect Nash equilibrium was to play the Nash equilibrium of
the stage game in every period
I In fact, one can prove generally that if the stage game has only one Nash
equilibrium then in the repeated game with that stage game, the unique subgame
perfect Nash equilibrium requires the Nash equilibrium to be played in all periods
and all information sets
I In contrast, in this game, we saw that there was a subgame perfect Nash
equilibrium in which an action profile (B1 , B2 ) that was not a Nash equilibrium of
the stage game was played in period 1
I What is the take away of this exercise?
I In the repeated Prisoner’s Dilemma, the stage game (played just once) had just
one Nash equilibrium
I The only subgame perfect Nash equilibrium was to play the Nash equilibrium of
the stage game in every period
I In fact, one can prove generally that if the stage game has only one Nash
equilibrium then in the repeated game with that stage game, the unique subgame
perfect Nash equilibrium requires the Nash equilibrium to be played in all periods
and all information sets
I In contrast, in this game, we saw that there was a subgame perfect Nash
equilibrium in which an action profile (B1 , B2 ) that was not a Nash equilibrium of
the stage game was played in period 1
I This was because there were multiple Nash equilibria of the stage game that
could be used as prize/punishment for certain behaviors
I Are there any other action profiles that can be played in the first period?
Normal Form

A2 B2 C2
A1 1, 1 0, 0 0, 0
B1 0, 0 4, 4 1, 5
C1 0, 0 5, 1 3, 3
I Are there any other action profiles that can be played in the first period?
Normal Form

A2 B2 C2
A1 1, 1 0, 0 0, 0
B1 0, 0 4, 4 1, 5
C1 0, 0 5, 1 3, 3

I Suppose that the players were to play (A1 , B2 ) in the first period
I Are there any other action profiles that can be played in the first period?
Normal Form

A2 B2 C2
A1 1, 1 0, 0 0, 0
B1 0, 0 4, 4 1, 5
C1 0, 0 5, 1 3, 3

I Suppose that the players were to play (A1 , B2 ) in the first period

I Can this occur? The answer is no

I Are there any other action profiles that can be played in the first period?
Normal Form

A2 B2 C2
A1 1, 1 0, 0 0, 0
B1 0, 0 4, 4 1, 5
C1 0, 0 5, 1 3, 3

I Suppose that the players were to play (A1 , B2 ) in the first period

I Can this occur? The answer is no

I Remember either (A1 , A2 ) or (C1 , C2 ) must be played in any pure strategy SPNE
after a history
I Now let us argue that (A1 , B2 ) cannot be played in period 1 in a SPNE
I Now let us argue that (A1 , B2 ) cannot be played in period 1 in a SPNE
I Suppose otherwise
I Now let us argue that (A1 , B2 ) cannot be played in period 1 in a SPNE
I Suppose otherwise
I No matter what happens in the second period, there is no way A1 could be a best
response against B2 in the first period.
I Now let us argue that (A1 , B2 ) cannot be played in period 1 in a SPNE
I Suppose otherwise
I No matter what happens in the second period, there is no way A1 could be a best
response against B2 in the first period.
I The maximum payoff that player 1 could get from playing according to this
“supposed” SPNE:
u1 (A1 , B2 ) + δu1 (C1 , C2 ) = 3δ
I Now let us argue that (A1 , B2 ) cannot be played in period 1 in a SPNE
I Suppose otherwise
I No matter what happens in the second period, there is no way A1 could be a best
response against B2 in the first period.
I The maximum payoff that player 1 could get from playing according to this
“supposed” SPNE:
u1 (A1 , B2 ) + δu1 (C1 , C2 ) = 3δ

I Now suppose that player 1 deviates to C1 instead of playing A1

I Now let us argue that (A1 , B2 ) cannot be played in period 1 in a SPNE
I Suppose otherwise
I No matter what happens in the second period, there is no way A1 could be a best
response against B2 in the first period.
I The maximum payoff that player 1 could get from playing according to this
“supposed” SPNE:
u1 (A1 , B2 ) + δu1 (C1 , C2 ) = 3δ

I Now suppose that player 1 deviates to C1 instead of playing A1

I The worst the payoff that he could get in any SPNE:

u1 (C1 , B2 ) + δu1 (A1 , A2 ) = 5 + δ

I Now let us argue that (A1 , B2 ) cannot be played in period 1 in a SPNE
I Suppose otherwise
I No matter what happens in the second period, there is no way A1 could be a best
response against B2 in the first period.
I The maximum payoff that player 1 could get from playing according to this
“supposed” SPNE:
u1 (A1 , B2 ) + δu1 (C1 , C2 ) = 3δ

I Now suppose that player 1 deviates to C1 instead of playing A1

I The worst the payoff that he could get in any SPNE:

u1 (C1 , B2 ) + δu1 (A1 , A2 ) = 5 + δ

I 5 + δ is always greater than 3δ

I Now let us argue that (A1 , B2 ) cannot be played in period 1 in a SPNE
I Suppose otherwise
I No matter what happens in the second period, there is no way A1 could be a best
response against B2 in the first period.
I The maximum payoff that player 1 could get from playing according to this
“supposed” SPNE:
u1 (A1 , B2 ) + δu1 (C1 , C2 ) = 3δ

I Now suppose that player 1 deviates to C1 instead of playing A1

I The worst the payoff that he could get in any SPNE:

u1 (C1 , B2 ) + δu1 (A1 , A2 ) = 5 + δ

I 5 + δ is always greater than 3δ

I By playing C1 against B2 , player 1 can guarantee a higher payoff
I Can there be a SPNE in which (A1 , C2 ) is played in period 1?
I Can there be a SPNE in which (A1 , C2 ) is played in period 1?

I The answer is no for the same reason

I Can there be a SPNE in which (A1 , C2 ) is played in period 1?

I The answer is no for the same reason

I By playing A1 against C2 , the best that player 1 can hope for in a SPNE is:

u1 (A1 , C2 ) + δu1 (C1 , C2 ) = 3δ

I Can there be a SPNE in which (A1 , C2 ) is played in period 1?

I The answer is no for the same reason

I By playing A1 against C2 , the best that player 1 can hope for in a SPNE is:

u1 (A1 , C2 ) + δu1 (C1 , C2 ) = 3δ

I The worst payoff that player 1 can obtain by playing C1 instead in period 1 is:

u1 (C1 , C2 ) + δu1 (A1 , A2 ) = 3 + δ

I Can there be a SPNE in which (A1 , C2 ) is played in period 1?

I The answer is no for the same reason

I By playing A1 against C2 , the best that player 1 can hope for in a SPNE is:

u1 (A1 , C2 ) + δu1 (C1 , C2 ) = 3δ

I The worst payoff that player 1 can obtain by playing C1 instead in period 1 is:

u1 (C1 , C2 ) + δu1 (A1 , A2 ) = 3 + δ

I 3 + δ is always greater than 3δ

I Can there be a SPNE in which (A1 , C2 ) is played in period 1?

I The answer is no for the same reason

I By playing A1 against C2 , the best that player 1 can hope for in a SPNE is:

u1 (A1 , C2 ) + δu1 (C1 , C2 ) = 3δ

I The worst payoff that player 1 can obtain by playing C1 instead in period 1 is:

u1 (C1 , C2 ) + δu1 (A1 , A2 ) = 3 + δ

I 3 + δ is always greater than 3δ

I Thus, there are incentives to deviate

I Symmetrically there cannot be any SPNE in which (B1 , A2 ) and (C1 , A2 ) are
played in period 1
I Symmetrically there cannot be any SPNE in which (B1 , A2 ) and (C1 , A2 ) are
played in period 1

I We already know that (A1 , A2 ), (B1 , B2 ), (C1 , C2 ) can be played in a SPNE in

period 1
I Symmetrically there cannot be any SPNE in which (B1 , A2 ) and (C1 , A2 ) are
played in period 1

I We already know that (A1 , A2 ), (B1 , B2 ), (C1 , C2 ) can be played in a SPNE in

period 1

I The remaining question is whether (C1 , B2 ) can be played in period 1

I Consider the following strategy profile

I Player 1’s strategy is:

1. Play C1 in period 1
2. Play A1 in period 2 if the first period action profile was (C1 , C2 )
3. Play C1 in period 2 if the first period action profile was anything other than (C1 , C2 )

I Player 2’s strategy is:

1. Play B2 in period 1
2. Play A2 in period 2 if the first period action profile was (C1 , C2 )
3. Play C2 in period 2 if the first period action profile was anything other than (C1 , C2 )
I We know that the strategy is a NE in the subgames that start in t = 2
I We know that the strategy is a NE in the subgames that start in t = 2

I But what about the whole game?

So we can simplify the game which gives the following game tree.

Anna

A1 C1
B1
Bob Bob Bob

A2 C2 A2 C2
B2 B2 A2 C2
B2
(1 + 3δ, 1 + 3δ) (3δ, 3δ) (3δ, 3δ) (1 + 3δ, 5 + 3δ)

(3δ, 3δ) (4 + 3δ, 4 + 3δ) (3δ, 3δ) (3 + δ, 3 + δ)

(5 + 3δ, 1 + 3δ)
The normal form of this game (conditional on what happens in T = 2) is:

Normal Form

A2 B2 C2
A1 1 + 3δ, 1 + 3δ 3δ, 3δ 3δ, 3δ
B1 3δ, 3δ 4 + 3δ, 4 + 3δ 1 + 3δ, 5 + 3δ
C1 3δ, 3δ 5 + 3δ, 1 + 3δ 3 + δ, 3 + δ
I In this game the best response for player i is:



 A1 if s2 = A2

C if
1 s2 = B2
BR1 (s2 ) =


 C 1 if s2 = C2

B if
1 s2 = C2 & δ = 1
I In this game the best response for player i is:



 A1 if s2 = A2

C if
1 s2 = B2
BR1 (s2 ) =


 C 1 if s2 = C2

B if
1 s2 = C2 & δ = 1

I In this game the best response for player 2 is:




 A2 if s1 = A1

C if s = B
2 1 1
BR2 (s1 ) =


 C 2 if s 1 = C 1

B if s = C & δ = 1
2 1 1
I In this game the best response for player i is:



 A1 if s2 = A2

C if
1 s2 = B2
BR1 (s2 ) =


 C 1 if s2 = C2

B if
1 s2 = C2 & δ = 1

I In this game the best response for player 2 is:




 A2 if s1 = A1

C if s = B
2 1 1
BR2 (s1 ) =


 C 2 if s 1 = C 1

B if s = C & δ = 1
2 1 1

I An equilibrium outcome of this game is to play (C1 , B2 ) in period 1 and (C1 , C2 )

in period 2 if δ = 1
I There are other SPNE that results in the same equilibrium outcome

I For example consider the following SPNE

I Player 1’s strategy is:

1. Play C1 in period 1.
2. Play A1 in period 2 if the first period action profile was anything other than (C1 , B2 ).
3. Play C1 in period 2 if the first period action profile was (C1 , B2 ).

1. Player 2’s strategy is:

2. Play B2 in period 1.
3. Play A2 in period 2 if the first period action profile was anything other than (C1 , B2 ).
4. Play C2 in period 2 if the first period action profile was (C1 , B2 ).
I We know that the strategy is a NE in the subgames that start in t = 2
I We know that the strategy is a NE in the subgames that start in t = 2

I But what about the whole game?

So we can simplify the game which gives the following game tree.

Anna

A1 C1
B1
Bob Bob Bob

A2 C2 A2 C2
B2 B2 A2 C2
B2
(1 + δ, 1 + δ) (δ, δ) (δ, δ) (1 + δ, 5 + δ)

(δ, δ) (4 + δ, 4 + δ) (δ, δ) (3 + δ, 3 + δ)

(5 + 3δ, 1 + 3δ)
The normal form of this game (conditional on what happens in T = 2) is:

Normal Form

A2 B2 C2
A1 1 + δ, 1 + δ 3δ, δ δ, δ
B1 δ, δ 4 + δ, 4 + δ 1 + δ, 5 + δ
C1 δ, δ 5 + 3δ, 1 + 3δ 3 + δ, 3 + δ
I In this game the best response for player i is:

A1 if s2 = A2

BR1 (s2 ) = C1 if s2 = B2

C1 if s2 = C2


I In this game the best response for player 2 is:


A2 if s1 = A1



C if s = B
2 1 1
BR2 (s1 ) =


 C 2 if s 1 = C 1

B if s = C & δ = 1
2 1 1

I An equilibrium outcome of this game is to play (C1 , B2 ) in period 1 and (C1 , C2 )

in period 2 if δ = 1
I There are many many pure strategy SPNE of this game!
I There are many many pure strategy SPNE of this game!

I The set of pure strategy SPNE can involve the play of non-stage game NE action
profiles in period 1 (although in period 2, players must play stage game NE)
I There are many many pure strategy SPNE of this game!

I The set of pure strategy SPNE can involve the play of non-stage game NE action
profiles in period 1 (although in period 2, players must play stage game NE)

I We’ve already seen that there may be multiple SPNE that lead to the same
equilibrium outcomes
I There are many many pure strategy SPNE of this game!

I The set of pure strategy SPNE can involve the play of non-stage game NE action
profiles in period 1 (although in period 2, players must play stage game NE)

I We’ve already seen that there may be multiple SPNE that lead to the same
equilibrium outcomes

I Thus, characterizing all pure strategy SPNE is extremely tedious

I There are many many pure strategy SPNE of this game!

I The set of pure strategy SPNE can involve the play of non-stage game NE action
profiles in period 1 (although in period 2, players must play stage game NE)

I We’ve already seen that there may be multiple SPNE that lead to the same
equilibrium outcomes

I Thus, characterizing all pure strategy SPNE is extremely tedious

I So instead of calculating all possible SPNE, lets just calculate the set of all
possible equilibrium outcomes
I We know that the following are possible equilibrium outcomes:
I We know that the following are possible equilibrium outcomes:
1. (A1 , A2 ), (A1 , A2 )
I We know that the following are possible equilibrium outcomes:
1. (A1 , A2 ), (A1 , A2 )
2. (A1 , A2 ), (C1 , C2 )
I We know that the following are possible equilibrium outcomes:
1. (A1 , A2 ), (A1 , A2 )
2. (A1 , A2 ), (C1 , C2 )
3. (C1 , C2 ), (A1 , A2 )
I We know that the following are possible equilibrium outcomes:
1. (A1 , A2 ), (A1 , A2 )
2. (A1 , A2 ), (C1 , C2 )
3. (C1 , C2 ), (A1 , A2 )
4. (C1 , C2 ), (C1 , C2 )
I We know that the following are possible equilibrium outcomes:
1. (A1 , A2 ), (A1 , A2 )
2. (A1 , A2 ), (C1 , C2 )
3. (C1 , C2 ), (A1 , A2 )
4. (C1 , C2 ), (C1 , C2 )
5. (B1 , B2 ), (C1 , C2 )
I We know that the following are possible equilibrium outcomes:
1. (A1 , A2 ), (A1 , A2 )
2. (A1 , A2 ), (C1 , C2 )
3. (C1 , C2 ), (A1 , A2 )
4. (C1 , C2 ), (C1 , C2 )
5. (B1 , B2 ), (C1 , C2 )
6. (C1 , B2 ), (C1 , C2 )
I We know that the following are possible equilibrium outcomes:
1. (A1 , A2 ), (A1 , A2 )
2. (A1 , A2 ), (C1 , C2 )
3. (C1 , C2 ), (A1 , A2 )
4. (C1 , C2 ), (C1 , C2 )
5. (B1 , B2 ), (C1 , C2 )
6. (C1 , B2 ), (C1 , C2 )
7. (B1 , C2 ), (C1 , C2 )
I We know that the following are possible equilibrium outcomes:
1. (A1 , A2 ), (A1 , A2 )
2. (A1 , A2 ), (C1 , C2 )
3. (C1 , C2 ), (A1 , A2 )
4. (C1 , C2 ), (C1 , C2 )
5. (B1 , B2 ), (C1 , C2 )
6. (C1 , B2 ), (C1 , C2 )
7. (B1 , C2 ), (C1 , C2 )

I Can there be other equilibrium outcomes?

I We know that the following are possible equilibrium outcomes:
1. (A1 , A2 ), (A1 , A2 )
2. (A1 , A2 ), (C1 , C2 )
3. (C1 , C2 ), (A1 , A2 )
4. (C1 , C2 ), (C1 , C2 )
5. (B1 , B2 ), (C1 , C2 )
6. (C1 , B2 ), (C1 , C2 )
7. (B1 , C2 ), (C1 , C2 )

I Can there be other equilibrium outcomes?

I We know that the following are possible equilibrium outcomes:
1. (A1 , A2 ), (A1 , A2 )
2. (A1 , A2 ), (C1 , C2 )
3. (C1 , C2 ), (A1 , A2 )
4. (C1 , C2 ), (C1 , C2 )
5. (B1 , B2 ), (C1 , C2 )
6. (C1 , B2 ), (C1 , C2 )
7. (B1 , C2 ), (C1 , C2 )

I Can there be other equilibrium outcomes? No!

I We know that the following are possible equilibrium outcomes:
1. (A1 , A2 ), (A1 , A2 )
2. (A1 , A2 ), (C1 , C2 )
3. (C1 , C2 ), (A1 , A2 )
4. (C1 , C2 ), (C1 , C2 )
5. (B1 , B2 ), (C1 , C2 )
6. (C1 , B2 ), (C1 , C2 )
7. (B1 , C2 ), (C1 , C2 )

I Can there be other equilibrium outcomes? No! Why?

Lecture 18: Repeated Games

Recap from last class

More than one NE in the stage game

Example 1

Example 2
Lecture 18: Repeated Games

Recap from last class

More than one NE in the stage game

Example 1

Example 2
Consider the following repeated game and δ = 1

Stage Game

A2 B2 C2
A1 (10, 10) (−1, 11) (−1, 11)
B1 (11, −1) (3, 1) (0, 0)
C1 (11, −1) (0, 0) (1, 3)
I The above game has two Nash equilibria (B1 , B2 ) and (C1 , C2 )
I The above game has two Nash equilibria (B1 , B2 ) and (C1 , C2 )

I Even though there are multiple Nash equilibria, there are no subgame perfect
equilibria in which (A1 , A2 ) is played in period 1
I The above game has two Nash equilibria (B1 , B2 ) and (C1 , C2 )

I Even though there are multiple Nash equilibria, there are no subgame perfect
equilibria in which (A1 , A2 ) is played in period 1

I Either (B1 , B2 ) or (C1 , C2 ) must be played after the history (A1 , A2 ) in period 1
since in the last period, always one of the stage game Nash equilibria must be
played.
Case 1:

I Suppose that (B1 , B2 ) is played in period 2 after (A1 , A2 ) in period 1

Case 1:

I Suppose that (B1 , B2 ) is played in period 2 after (A1 , A2 ) in period 1

I Player 2 obtains a payoff of

10 + δ
Case 1:

I Suppose that (B1 , B2 ) is played in period 2 after (A1 , A2 ) in period 1

I Player 2 obtains a payoff of

10 + δ

I By deviating to B2 in period 1, player 2 obtains at least:

11 + δ

since in period 2 either (B1 , B2 ) or (C1 , C2 ) will be played in any SPNE

Case 1:

I Suppose that (B1 , B2 ) is played in period 2 after (A1 , A2 ) in period 1

I Player 2 obtains a payoff of

10 + δ

I By deviating to B2 in period 1, player 2 obtains at least:

11 + δ

since in period 2 either (B1 , B2 ) or (C1 , C2 ) will be played in any SPNE

I Thus there are incentives to deviate

Case 2:

I Suppose instead that (C1 , C2 ) is played in period 2 after (A1 , A2 ) in period 1

Case 2:

I Suppose instead that (C1 , C2 ) is played in period 2 after (A1 , A2 ) in period 1

I player 1 obtains a payoff of

10 + δ
Case 2:

I Suppose instead that (C1 , C2 ) is played in period 2 after (A1 , A2 ) in period 1

I player 1 obtains a payoff of

10 + δ

I By deviating to B1 in period 1, player 1 obtains at least 11 + δ

Case 2:

I Suppose instead that (C1 , C2 ) is played in period 2 after (A1 , A2 ) in period 1

I player 1 obtains a payoff of

10 + δ

I By deviating to B1 in period 1, player 1 obtains at least 11 + δ

I Thus there are incentives to deviate

I Even though there are multiple NE in the stage game, it may still be impossible to
achieve Pareto efficient action profiles in period 1
I Even though there are multiple NE in the stage game, it may still be impossible to
achieve Pareto efficient action profiles in period 1

I The key to this example was that players disagreed on which stage game NE is
better
I Even though there are multiple NE in the stage game, it may still be impossible to
achieve Pareto efficient action profiles in period 1

I The key to this example was that players disagreed on which stage game NE is
better

I Thus, at least one person always had an incentive to deviate away from (A1 , A2 )
in period 1
Lecture 18: Repeated Games

Recap from last class

More than one NE in the stage game

Example 1

Example 2
Lecture 18: Repeated Games

Recap from last class

More than one NE in the stage game

Example 1

Example 2
I Even if there is disagreement about which stage game NE is better between the
two players, we can still obtain examples of outcomes that are not Nash
equilibrium in the first period
I Even if there is disagreement about which stage game NE is better between the
two players, we can still obtain examples of outcomes that are not Nash
equilibrium in the first period

I Consider for example the following stage game and suppose we consider a twice
repeated game with discount factor δ > 12

Stage Game

A2 B2 C2
A1 (10, 10) (0, 9) (0, 9)
B1 (11, −1) (3, 1) (0, 0)
C1 (11, −2) (0, 0) (1, 3)
I The NE of the stage game are (B1 , B2 ) and (C1 , C2 )
I The NE of the stage game are (B1 , B2 ) and (C1 , C2 )

I In this repeated game, is there a subgame perfect Nash equilibrium in which

(A1 , A2 ) is played in period 1?
I The NE of the stage game are (B1 , B2 ) and (C1 , C2 )

I In this repeated game, is there a subgame perfect Nash equilibrium in which

(A1 , A2 ) is played in period 1?

I The answer is yes

I Consider the following strategy profile

I Player 1 plays the following strategy:

1. A1 in period 1;
2. B1 in period 2 if (A1 , A2 ) was played in period 1;
3. C1 in period 2 if (A1 , A2 ) was not played in period 1.

I Player 2 plays the following strategy:

1. A2 in period 1;
2. B2 in period 2 if (A1 , A2 ) was played in period 1;
3. C2 in period 2 if (A1 , A2 ) was not played in period 1.
I Is the above an SPNE?
I Is the above an SPNE?

I no (if δ < 12 )!

Stage Game

A2 B2 C2
A1 (10, 10) (0, 9) (0, 9)
B1 (11, −1) (3, 1) (0, 0)
C1 (11, −2) (0, 0) (1, 3)
I Is the above an SPNE?

I no (if δ < 12 )!

Stage Game

A2 B2 C2
A1 (10, 10) (0, 9) (0, 9)
B1 (11, −1) (3, 1) (0, 0)
C1 (11, −2) (0, 0) (1, 3)

I Player 1:
I Is the above an SPNE?

I no (if δ < 12 )!

Stage Game

A2 B2 C2
A1 (10, 10) (0, 9) (0, 9)
B1 (11, −1) (3, 1) (0, 0)
C1 (11, −2) (0, 0) (1, 3)

I Player 1:

I If he follows: u1 = 10 + 3δ
I Is the above an SPNE?

I no (if δ < 12 )!

Stage Game

A2 B2 C2
A1 (10, 10) (0, 9) (0, 9)
B1 (11, −1) (3, 1) (0, 0)
C1 (11, −2) (0, 0) (1, 3)

I Player 1:

I If he follows: u1 = 10 + 3δ

I If he defects: u1 = 11 + δ
I Is the above an SPNE?

I no (if δ < 12 )!

Stage Game

A2 B2 C2
A1 (10, 10) (0, 9) (0, 9)
B1 (11, −1) (3, 1) (0, 0)
C1 (11, −2) (0, 0) (1, 3)

I Player 1:

I If he follows: u1 = 10 + 3δ

I If he defects: u1 = 11 + δ

I Follows if δ ≥ 1
2
I Player 2:
I Player 2:

I If he follows: u2 = 10 + δ
I Player 2:

I If he follows: u2 = 10 + δ

I If he defects: u2 = 9 + 3δ
I Player 2:

I If he follows: u2 = 10 + δ

I If he defects: u2 = 9 + 3δ

I Follows if δ ≤ 1
2
I Player 2:

I If he follows: u2 = 10 + δ

I If he defects: u2 = 9 + 3δ

I Follows if δ ≤ 1
2

I Can only be a SPNE is δ = 1

2
I The key here is that player 2 by breaking the agreement in period 1 moves the
period 2 play to his favored stage game NE of (C1 , C2 )
I Suppose we flipped the roles of B and C and considered the following strategy
profile

I Player 1 plays the following strategy:

1. A1 in period 1;
2. C1 in period 2 if (A1 , A2 ) was played in period 1;
3. B1 in period 2 if (A1 , A2 ) was not played in period 1.

I Player 2 plays the following strategy:

1. A2 in period 1;
2. C2 in period 2 if (A1 , A2 ) was played in period 1;
3. B2 in period 2 if (A1 , A2 ) was not played in period 1.
I This is not a SPNE either because now player 1 has a definitive incentive to
deviate from (A1 , A2 ) in period 1
Stage Game

A2 B2 C2
A1 (10, 10) (0, 9) (0, 9)
B1 (11, −1) (3, 1) (0, 0)
C1 (11, −2) (0, 0) (1, 3)
I This is not a SPNE either because now player 1 has a definitive incentive to
deviate from (A1 , A2 ) in period 1
Stage Game

A2 B2 C2
A1 (10, 10) (0, 9) (0, 9)
B1 (11, −1) (3, 1) (0, 0)
C1 (11, −2) (0, 0) (1, 3)

I Player 1:
I This is not a SPNE either because now player 1 has a definitive incentive to
deviate from (A1 , A2 ) in period 1
Stage Game

A2 B2 C2
A1 (10, 10) (0, 9) (0, 9)
B1 (11, −1) (3, 1) (0, 0)
C1 (11, −2) (0, 0) (1, 3)

I Player 1:

I If he follows: u1 = 10 + δ
I This is not a SPNE either because now player 1 has a definitive incentive to
deviate from (A1 , A2 ) in period 1
Stage Game

A2 B2 C2
A1 (10, 10) (0, 9) (0, 9)
B1 (11, −1) (3, 1) (0, 0)
C1 (11, −2) (0, 0) (1, 3)

I Player 1:

I If he follows: u1 = 10 + δ

I If he defects: u1 = 11 + 3δ
I This is not a SPNE either because now player 1 has a definitive incentive to
deviate from (A1 , A2 ) in period 1
Stage Game

A2 B2 C2
A1 (10, 10) (0, 9) (0, 9)
B1 (11, −1) (3, 1) (0, 0)
C1 (11, −2) (0, 0) (1, 3)

I Player 1:

I If he follows: u1 = 10 + δ

I If he defects: u1 = 11 + 3δ

I Always defects
I So how do we construct a SPNE with (A1 , A2 ) played in period 1?
I So how do we construct a SPNE with (A1 , A2 ) played in period 1?

I The key here is to notice that player 2 does not need to be punished in period 2
from breaking the agreement in period 1
I So how do we construct a SPNE with (A1 , A2 ) played in period 1?

I The key here is to notice that player 2 does not need to be punished in period 2
from breaking the agreement in period 1

I This is because in period 1 player 2 is best responding myopically at (A1 , A2 )

already
I So how do we construct a SPNE with (A1 , A2 ) played in period 1?

I The key here is to notice that player 2 does not need to be punished in period 2
from breaking the agreement in period 1

I This is because in period 1 player 2 is best responding myopically at (A1 , A2 )

already

I In other words, need to be punished only if the player has a deviation that
benefits him myopically or in the short term
I Player 1 plays the following strategy:
1. A1 in period 1;
2. B1 in period 2 if player 1 played A1 ;
3. C1 in period 2 if player 1 played B1 or C1 .

I Player 2 plays the following strategy:

1. A2 in period 1;
2. B2 in period 2 if player 1 played A1 ;
3. C2 in period 2 if player 1 played B1 or C1 .
 Stage Game

A2 B2 C2
A1 (10, 10) (0, 9) (0, 9)
B1 (11, −1) (3, 1) (0, 0)
C1 (11, −2) (0, 0) (1, 3)

I Player 1:
 Stage Game

A2 B2 C2
A1 (10, 10) (0, 9) (0, 9)
B1 (11, −1) (3, 1) (0, 0)
C1 (11, −2) (0, 0) (1, 3)

I Player 1:
I If he follows: u1 = 10 + 3δ
 Stage Game

A2 B2 C2
A1 (10, 10) (0, 9) (0, 9)
B1 (11, −1) (3, 1) (0, 0)
C1 (11, −2) (0, 0) (1, 3)

I Player 1:
I If he follows: u1 = 10 + 3δ
I If he defects: u1 = 11 + δ
 Stage Game

A2 B2 C2
A1 (10, 10) (0, 9) (0, 9)
B1 (11, −1) (3, 1) (0, 0)
C1 (11, −2) (0, 0) (1, 3)

I Player 1:
I If he follows: u1 = 10 + 3δ
I If he defects: u1 = 11 + δ
I Follows if δ ≥ 12
 Stage Game

A2 B2 C2
A1 (10, 10) (0, 9) (0, 9)
B1 (11, −1) (3, 1) (0, 0)
C1 (11, −2) (0, 0) (1, 3)

I Player 1:
I If he follows: u1 = 10 + 3δ
I If he defects: u1 = 11 + δ
I Follows if δ ≥ 12

I Player 2:
 Stage Game

A2 B2 C2
A1 (10, 10) (0, 9) (0, 9)
B1 (11, −1) (3, 1) (0, 0)
C1 (11, −2) (0, 0) (1, 3)

I Player 1:
I If he follows: u1 = 10 + 3δ
I If he defects: u1 = 11 + δ
I Follows if δ ≥ 12

I Player 2:
I If he follows: u2 = 10 + X δ
 Stage Game

A2 B2 C2
A1 (10, 10) (0, 9) (0, 9)
B1 (11, −1) (3, 1) (0, 0)
C1 (11, −2) (0, 0) (1, 3)

I Player 1:
I If he follows: u1 = 10 + 3δ
I If he defects: u1 = 11 + δ
I Follows if δ ≥ 12

I Player 2:
I If he follows: u2 = 10 + X δ
I If he defects: u2 = 9 + X δ
 Stage Game

A2 B2 C2
A1 (10, 10) (0, 9) (0, 9)
B1 (11, −1) (3, 1) (0, 0)
C1 (11, −2) (0, 0) (1, 3)

I Player 1:
I If he follows: u1 = 10 + 3δ
I If he defects: u1 = 11 + δ
I Follows if δ ≥ 12

I Player 2:
I If he follows: u2 = 10 + X δ
I If he defects: u2 = 9 + X δ
I Follows