Microeconomic Analysis (806L1)

Topic 3: Game Theory

Dynamic Games

Matthew Embrey

University of Sussex

Autumn 2023

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 1 / 23

Dynamic Games
Extensive Form

In the extensive form, explicitly consider

the timing of actions
the information players have when they are called upon to act
Represented by game trees
nodes denoting players including possibly Nature,
branches denoting choices, and
payoffs at terminal nodes
(unique start node; every node has a unique path to start node).
1
L R

2 2
L′ R′ L′ R′

(3, 1) (1, 2) (2, 1) (0, 0)

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 2 / 23

Dynamic Games
Extensive to Strategic Form

To every extensive form game, there is a corresponding strategic

(normal) form.
Thus far, in the strategic form games we looked at, strategies were
simple choices e.g. H or T in Matching Pennies.
For extensive form games, think of a strategy as a complete contingent
plan of action: it should prescribe an action for every place a player
might be called upon to act.
A given strategic form game can generally correspond to several
extensive form games.

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 3 / 23

Dynamic Games
Extensive to Strategic Form

1
L R

2 2
L′ R′ L′ R′

(3, 1) (1, 2) (2, 1) (0, 0)

Firstly, what strategies do the two players have? L or R for player 1.

How about for player 2?

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 4 / 23

Dynamic Games
Extensive to Strategic Form

1
L R

2 2
L′ R′ L′ R′

(3, 1) (1, 2) (2, 1) (0, 0)

Need to specify a choice for P2 for every node; e.g., (L’, L’) means
choose L’ should P1 choose L and L′ should P1 choose R.
P2
(L’, L’) (L’, R’) (R’, L’) (R’, R’)
P1 L
R

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 4 / 23

Dynamic Games
Extensive to Strategic Form

1
L R
y

2 2
L′ R′ L′ R′



(3,1) (1, 2) (2, 1) (0, 0)

Follow the arrows to find the payoffs for each cell...

P2
(L’, L’) (L’, R’) (R’, L’) (R’, R’)
P1 L (3,1)
R

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 4 / 23

Dynamic Games
Extensive to Strategic Form

1
L R
y

2 2
L′ R′ L′ R′


(3,1) (1, 2) (2, 1) (0, 0)

Follow the arrows to find the payoffs for each cell...

P2
(L’, L’) (L’, R’) (R’, L’) (R’, R’)
P1 L (3,1) (3,1)
R

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 4 / 23

Dynamic Games
Extensive to Strategic Form

1
L R
y

2 2
L′ R′ L′ R′


(3, 1) (1,2) (2, 1) (0, 0)

Follow the arrows to find the payoffs for each cell...

P2
(L’, L’) (L’, R’) (R’, L’) (R’, R’)
P1 L (3,1) (3,1) (1,2)
R

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 4 / 23

Dynamic Games
Extensive to Strategic Form

1
L R
y

2 2
L′ R′ L′ R′
 
(3, 1) (1,2) (2, 1) (0, 0)

Follow the arrows to find the payoffs for each cell...

P2
(L’, L’) (L’, R’) (R’, L’) (R’, R’)
P1 L (3,1) (3,1) (1,2) (1,2)
R

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 4 / 23

Dynamic Games
Extensive to Strategic Form

1
L R
%
2 2
L′ R′ L′ R′



(3, 1) (1, 2) (2,1) (0, 0)

Follow the arrows to find the payoffs for each cell...

P2
(L’, L’) (L’, R’) (R’, L’) (R’, R’)
P1 L (3,1) (3,1) (1,2) (1,2)
R (2,1)

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 4 / 23

Dynamic Games
Extensive to Strategic Form

1
L R
%
2 2
L′ R′ L′ R′


(3, 1) (1, 2) (2, 1) (0,0)

Follow the arrows to find the payoffs for each cell...

P2
(L’, L’) (L’, R’) (R’, L’) (R’, R’)
P1 L (3,1) (3,1) (1,2) (1,2)
R (2,1) (0,0)

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 4 / 23

Dynamic Games
Extensive to Strategic Form

1
L R
%
2 2
L′ R′ L′ R′


(3, 1) (1, 2) (2,1) (0, 0)

Follow the arrows to find the payoffs for each cell...

P2
(L’, L’) (L’, R’) (R’, L’) (R’, R’)
P1 L (3,1) (3,1) (1,2) (1,2)
R (2,1) (0,0) (2,1)

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 4 / 23

Dynamic Games
Extensive to Strategic Form

1
L R
%
2 2
L′ R′ L′ R′
 
(3, 1) (1, 2) (2, 1) (0,0)

Follow the arrows to find the payoffs for each cell...

P2
(L’, L’) (L’, R’) (R’, L’) (R’, R’)
P1 L (3,1) (3,1) (1,2) (1,2)
R (2,1) (0,0) (2,1) (0,0)

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 4 / 23

Dynamic Games
Simultaneous Moves: Information Sets

Simultaneous moves or incomplete information can be represented in

the extensive form using information sets.
An information set is a collection of decision nodes for a player such
that she cannot distinguish between the nodes.

1
P2
L’ R’ L R

P1 L (3,1) (1,2) 2 2
R (2,1) (0,0) L′ R′ L′ R′

(3, 1) (1, 2) (2, 1) (0, 0)

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 5 / 23

Subgame Perfect Nash Equilibrium
The Backward Induction Algorithm

1
L R

2 2
L′ R′ L′ R′

(3, 1) (1, 2) (2, 1) (0, 0)

Start at the end of the tree - at any ultimate node, the player will
pick an action that yields the highest payoff to herself.

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 6 / 23

Subgame Perfect Nash Equilibrium
The Backward Induction Algorithm

2 2
L′ R′ L′ R′


(3, 1) (1, 2) (2, 1) (0, 0)

Start at the end of the tree - at any ultimate node, the player will
pick an action that yields the highest payoff to herself.

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 6 / 23

Subgame Perfect Nash Equilibrium
The Backward Induction Algorithm

1
L R
%
2 2
R′ L′


(1, 2) (2, 1)
Start at the end of the tree - at any ultimate node, the player will
pick an action that yields the highest payoff to herself.
These payoffs become the relevant payoffs for choices at the
penultimate nodes and one can work backwards to solve for the
backward induction outcome(s).

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 6 / 23

Subgame Perfect Nash Equilibrium
The Backward Induction Algorithm

1
L R
%
2 2
L′ R′ L′ R′


(3, 1) (1, 2) (2,1) (0, 0)
Start at the end of the tree - at any ultimate node, the player will
pick an action that yields the highest payoff to herself.
These payoffs become the relevant payoffs for choices at the
penultimate nodes and one can work backwards to solve for the
backward induction outcome(s).
Every game of complete and perfect information (information sets are
singleton nodes) with a finite number of nodes can be solved by
backward induction.
Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 6 / 23
Subgame Perfect Nash Equilibrium
Subgames

1
L R

2 2
L′ R′ L′ R′

(3, 1) (1, 2) (2, 1) (0, 0)

Subgame: A smaller piece of a game that can be regarded as a

game in its own right. It starts from a single node and all players at
that point know all of the past history of the game.
The above game has 3 sub-games (including the game itself).

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 7 / 23

Subgame Perfect Nash Equilibrium
Subgames

1
L R

2 2
L′ R′ L′ R′

(3, 1) (1, 2) (2, 1) (0, 0)

Subgame: A smaller piece of a game that can be regarded as a

game in its own right. It starts from a single node and all players at
that point know all of the past history of the game.
The above game has 3 sub-games (including the game itself).
Subgame Perfect Nash Equilibrium (SPNE): is a NE where the
players’ strategies constitute a NE in each sub-game.

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 7 / 23

Subgame Perfect Nash Equilibrium
Subgames

1
L R

2 2
L′ R′ L′ R′

(3, 1) (1, 2) (2, 1) (0, 0)

Subgame: A smaller piece of a game that can be regarded as a

game in its own right. It starts from a single node and all players at
that point know all of the past history of the game.
The above game has 3 sub-games (including the game itself).
Subgame Perfect Nash Equilibrium (SPNE): is a NE where the
players’ strategies constitute a NE in each sub-game.
Try finding the SPNE for a random extensive form game...
Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 7 / 23
Subgame Perfect Nash Equilibrium
Comparison to Nash Equilibrium

P2
(L’, L’) (L’, R’) (R’, L’) (R’, R’)
P1 L (3, 1) (3, 1) (1, 2) (1, 2)
R (2, 1) (0, 0) (2, 1) (0, 0)

The pure strategy Nash equilibrium are (R, (R’,L’)) and (L, (R’,R’)).
Recall that the backward induction outcome was (R, L’). The SPNE
is (R, (R’,L’)).
What about (L, (R’,R’))? It is a NE. Here player 2 plans to play R’ if
player 1 chose L, and also if player 1 chose R. But is that credible? If
player 1 did chose R, player 2’s best response is L’.

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 8 / 23

Subgame Perfect Nash Equilibrium
Importance of Off-Equilibrium Play

Backward induction outcome is

1
L R

(1, 0) 2
L R

(0, 1) 1
t b

(4, 0) (2, 2)

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 9 / 23

Subgame Perfect Nash Equilibrium
Importance of Off-Equilibrium Play

Backward induction outcome is

1 P1 plays L and payoffs are (1,0).
The rest of the tree is not reached
L R in equilibrium play.

(1,0) 2 But 1’s strategy should still be
complete.
L R
~ He has 4 strategies: Lt, Lb, Rt,
(0, 1) 1 and Rb.
What player 1 will do at his 2nd
t b node determines what player 2 will


do in his turn which in turn causes
(4, 0) (2, 2)
1 to choose L.
SPNE is (Lt, L).

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 9 / 23

Repeated games
Example One-Shot Game

Cooperate Defect
Cooperate (1, 1) (-1, 2)
Defect (2, -1) (0, 0)

Recall the Prisoner’s dilemma

A workhorse (stage) game for social dilemmas
‘Defect’ strictly dominates ‘Cooperate’; (D,D) is the unique NE which
is Pareto dominated by the cooperative outcome.
But in many such situations in real life, we observe some cooperation.
Players often interact repeatedly over time
game is not “one-shot”
can this foster cooperation?

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 10 / 23

Repeated games
Repetition of the same Stage Game

Repeated games: same stage game is played each period for T

periods (perhaps T → ∞)
A special class of dynamic game
a game of imperfect (simultaneous-moves during a stage) but complete
information (no move of nature)
repetition of the same stage game
We discount future payoffs because of impatience, uncertainty etc.
Discount rate, δ ∈ [0, 1)
Total (discounted) payoff for player i in a T period game:
PT
Ui = t=0 δ t uit where uit is i ′ s payoff in the stage game in period t.
Useful formula for sum of infinite geometric series:
1
S = 1 + δ + δ 2 + δ 3 + ... = 1−δ

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 11 / 23

Repeated games
Finitely repeated Prisoner’s Dilemma

Players often interact for T < ∞ periods.

What is the sub game perfect Nash equilibrium?

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 12 / 23

Repeated games
Finitely repeated Prisoner’s Dilemma

Players often interact for T < ∞ periods.

What is the sub game perfect Nash equilibrium?
Backward induction - reason backwards from the final period.
Defect is dominant in the last period, so the sub game at T has (D,D)
as the NE.
Since period T play is (D,D), the play at T − 1 will also be (D,D)
regardless of the history of play until then.
Unique SPE: (D,D) in each period.
In general, for any finitely repeated game in which the stage game has
a unique pure strategy equilibrium a∗ , there is a unique SPNE with a∗
played in each period t = 0, 1, 2, ...T .

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 12 / 23

Repeated games
Cooperation in finitely repeated games

No cooperation in finitely repeated games with unique stage

equilibrium.
Suppose the stage game has multiple equilibria:
X Y Z
X (3, 3) (0, 4) (-2, 0)
Y (4, 0) (1, 1) (-2, 0)
Z (0, -2) (0, -2) (-1, -1)

(Y,Y) and (Z,Z) are the two pure strategy NE. Consider a 2 period
repeated game (no discounting):
Strategy: play (X,X) in period 1 and (Y,Y) in period 2. Punish a
deviation in period 1 by playing (Z,Z) in period 2.

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 13 / 23

Repeated games
Cooperation in finitely repeated games

No cooperation in finitely repeated games with unique stage

equilibrium.
Suppose the stage game has multiple equilibria:
X Y Z
X (3, 3) (0, 4) (-2, 0)
Y (4, 0) (1, 1) (-2, 0)
Z (0, -2) (0, -2) (-1, -1)

(Y,Y) and (Z,Z) are the two pure strategy NE. Consider a 2 period
repeated game (no discounting):
Strategy: play (X,X) in period 1 and (Y,Y) in period 2. Punish a
deviation in period 1 by playing (Z,Z) in period 2.
If a player sticks to strategy, she gets 3 + 1 = 4.
If she deviates to playing Y in period 1, she knows that the other player
will punish her by playing Z in period 2, and so she will also play Z.
Payoff=4 − 1 = 3 so deviation is unprofitable.
The threat of (Z,Z) in the 2nd period supports cooperation.
Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 13 / 23
Repeated games
Infinitely repeated Prisoner’s Dilemma

Strategies are tricky to define in infinitely repeated games:

Recall that a strategy is a complete contingent plan of action.
Has to specify an action for a player in each period, contingent on the
history of play.
Many many possible histories. e.g one possible history in period 3 of
repeated Prisoner’s Dilemma , h3 =((C,C), (D,C), (D,D)).
Let us consider a trigger strategy, s ∗ in Prisoner’s Dilemma:
Play C in every period unless someone has ever played D in the past.
Play D forever if someone has ever played D in the past.
Note that s ∗ is a strategy since it prescribes an action for each period
contingent on all possible histories.
Key idea: support cooperation with the threat of (grim) punishment.

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 14 / 23

Repeated games
Infinitely repeated Prisoner’s Dilemma

Cooperate Defect
Cooperate (1, 1) (-1, 2)
Defect (2, -1) (0, 0)

Is (s ∗ , s ∗ ) SPNE?

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 15 / 23

Repeated games
Infinitely repeated Prisoner’s Dilemma

Cooperate Defect
Cooperate (1, 1) (-1, 2)
Defect (2, -1) (0, 0)

Is (s ∗ , s ∗ ) SPNE?
First, show that cooperation is best response to cooperation to date.

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 15 / 23

Repeated games
Infinitely repeated Prisoner’s Dilemma

Cooperate Defect
Cooperate (1, 1) (-1, 2)
Defect (2, -1) (0, 0)

Is (s ∗ , s ∗ ) SPNE?
First, show that cooperation is best response to cooperation to date.
Suppose no one has played D so far. If the other player is playing s ∗
1
Payoff from cooperation: 1 + δ + δ 2 + ... = 1−δ
Payoff from defection: 2 + 0 + 0 + ... = 2
1
Cooperation yields more if 1−δ ≥ 2 ⇒ δ ≥ 0.5

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 15 / 23

Repeated games
Infinitely repeated Prisoner’s Dilemma

Cooperate Defect
Cooperate (1, 1) (-1, 2)
Defect (2, -1) (0, 0)

Is (s ∗ , s ∗ ) SPNE?
First, show that cooperation is best response to cooperation to date.
Suppose no one has played D so far. If the other player is playing s ∗
1
Payoff from cooperation: 1 + δ + δ 2 + ... = 1−δ
Payoff from defection: 2 + 0 + 0 + ... = 2
1
Cooperation yields more if 1−δ ≥ 2 ⇒ δ ≥ 0.5
Second, defection is best response to any defection in the past.

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 15 / 23

Repeated games
Infinitely repeated Prisoner’s Dilemma

Cooperate Defect
Cooperate (1, 1) (-1, 2)
Defect (2, -1) (0, 0)

Is (s ∗ , s ∗ ) SPNE?
First, show that cooperation is best response to cooperation to date.
Suppose no one has played D so far. If the other player is playing s ∗
1
Payoff from cooperation: 1 + δ + δ 2 + ... = 1−δ
Payoff from defection: 2 + 0 + 0 + ... = 2
1
Cooperation yields more if 1−δ ≥ 2 ⇒ δ ≥ 0.5
Second, defection is best response to any defection in the past.
According to s ∗ , the other player will play D.
Then D is the best response.

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 15 / 23

Repeated games
Infinitely repeated Prisoner’s Dilemma

Cooperate Defect
Cooperate (1, 1) (-1, 2)
Defect (2, -1) (0, 0)

Is (s ∗ , s ∗ ) SPNE?
First, show that cooperation is best response to cooperation to date.
Suppose no one has played D so far. If the other player is playing s ∗
1
Payoff from cooperation: 1 + δ + δ 2 + ... = 1−δ
Payoff from defection: 2 + 0 + 0 + ... = 2
1
Cooperation yields more if 1−δ ≥ 2 ⇒ δ ≥ 0.5
Second, defection is best response to any defection in the past.
According to s ∗ , the other player will play D.
Then D is the best response.
Therefore, (s ∗ , s ∗ ) is SPNE for δ ≥ 0.5.
Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 15 / 23
Repeated games
Folk theorem(s)

The SPNE strategy we just discussed is the grim trigger strategy -

cooperation is supported by the threat of punishment forever.
Many other strategy profiles can be NE and/or SPNE.
Folk theorems: If players are sufficiently patient (δ is high enough),
any feasible individually rational set of payoffs can be supported by an
equilibrium.
Feasible payoffs are convex combinations of the pure strategy payoffs.
An individually rational payoff is a payoff that is greater than a player’s
reservation utility i.e. the payoff a player can guarantee herself in any
equilibrium (minimax payoff.)
Another example of a Nash equilibrium is Tit-for-tat: cooperate in
the starting period, and then do what the other player did in the
preceding period.

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 16 / 23

Bargaining
Cooperative versus Non-Cooperative Approach

Cooperative (or Axiomatic) approach:

Look for the subset of feasible outcomes that have ‘reasonable’
properties e.g. efficiency, symmetric, independent of irrelevant
alternatives etc.
Not concerned with the process of bargaining.
Most famous: Nash bargaining solution.
Noncooperative approach:
Models the bargaining process.
We will study the alternative offers bargaining game between two
players, i = 1, 2.
Players make sequential offers and counter-offers to split £1 until one
player accepts.
If game ends without acceptance, both players get 0.
Payoffs are discounted in each successive round , δi ∈ [0, 1]
We will study the finite horizon and the infinite horizon game.

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 17 / 23

The Ultimatum Game
Take-it or Leave-it Offer

Two players - a proposer and a responder.

Proposer is given £1000.
She makes an offer to the responder (£x) and keep the rest
(£1000-x.)
If responder accepts, money is split accordingly.
If responder rejects, each gets 0.
What are the NE? SPNE?

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 18 / 23

The Ultimatum Game
Subgame Perfect Nash Equilibrium

Consider the following strategy pair:

Proposer offers £300;
Responder rejects any offer less than 300, and accepts any offer
x ≥ 300.

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 19 / 23

The Ultimatum Game
Subgame Perfect Nash Equilibrium

Consider the following strategy pair:

Proposer offers £300;
Responder rejects any offer less than 300, and accepts any offer
x ≥ 300.
This is a Nash equilibrium: neither has a unilateral incentive to
deviate.

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 19 / 23

The Ultimatum Game
Subgame Perfect Nash Equilibrium

Consider the following strategy pair:

Proposer offers £300;
Responder rejects any offer less than 300, and accepts any offer
x ≥ 300.
This is a Nash equilibrium: neither has a unilateral incentive to
deviate.
But this is not SPNE - the responder’s strategy involves refusing
positive offers which is not credible.
The only SPNE has the proposer making the smallest possible
positive offer, and the responder accepting all positive offers.

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 19 / 23

Alternative Offers Bargaining
Finite Horizon

Consider the 2 period version with a common discount factor (δ).

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 20 / 23

Alternative Offers Bargaining
Finite Horizon

Consider the 2 period version with a common discount factor (δ).

Solve using backward induction:
Node (iii): 1 will accept any offer such that δ(1 − y ) ≥ 0 ⇒ y ≤ 1.
Assume players accept if indifferent. Then 2 will set y = 1.
Node (ii): 2 is assured of at least δy = δ if the game reaches node (iii).
So, 2 will accept only if (1 − x) ≥ δ ⇒ x ≤ 1 − δ.
Node (i): 1 will offer (1 − δ, δ) and 2 will accept.

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 20 / 23

Alternative Offers Bargaining
Finite Horizon

Unique SPNE for the 2-period case:

Player 2: In period 1, accept any offer such that 1 − x ≥ δ, else reject
and propose split (0,1) in period 2.
Player 1: In period 1, offer split (1 − δ, δ), and in period 2 accept any
offer.
The predicted outcome is efficient - game ends in the first period (no
discounting).
Last mover advantage especially if δ is high, i.e. players are patient.
Can be easily generalised to T > 2 periods and different discount
factors:
Last mover advantage gets muted as T increases.
More patient player has an edge.

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 21 / 23

Alternative Offers Bargaining
Infinite Horizon

Let player 1 be the first to offer; discount factors are δi , i = 1, 2.

Let M be the maximum possible share that player 1 can obtain
among all possible perfect equilibria in a period t when she makes an
offer.
In period t − 1, player 2 has to offer player 1 at least δ1 M to induce
player 1 to accept. Player 2 is assured of a minimum 1 − δ1 M for
herself.
In period t − 2, player 1 has to offer at least the discounted value of
the minimum that player 2 can assure herself in the next period i.e.
δ2 (1 − δ1 M). Therefore the best player 1 can do in this period is
1 − δ2 (1 − δ1 M).
The best that player 1 can do in period t − 2 is M by our definition:
1−δ2
M = 1 − δ2 (1 − δ1 M) ⇒ M = 1−δ 1 δ2

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 22 / 23

Alternative Offers Bargaining
Infinite Horizon

We will next solve for m which is the minimum share of player 1 in

all possible equilibria when she makes an offer in period t.
In period t − 1, since player 1 is assured of δ1 m, player 2 can keep a
maximum of 1 − δ1 m for herself.
In period t − 2, player 1 knows that player 2 will accept δ2 (1 − δ1 m).
Therefore at a minimum player 1 can get 1 − δ2 (1 − δ1 m).
The minimum that player can do in period t − 2 is m by our
1−δ2
definition: m = 1 − δ2 (1 − δ1 m) ⇒ m = 1−δ 1 δ2
M = m, hence there is a unique SPNE.
Note that for a common discount factor, player 1’s share becomes
1
1+δ .

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 23 / 23