MiA T2 StaticGames

Microeconomic Analysis (806L1)
Topic 2: Game Theory

Static Games
Matthew Embrey
University of Sussex
Autumn 2023
Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 1 / 31

Strategic Interaction
Why Game Theory?
“No man is an island.”

Interdependence is a central feature of real life; one’s best choice
depends on what others do or might do, and vice versa.
Game theory is the formal modelling of such strategic interactions.
Game theory has become an indispensable tool of economic analysis.
Routinely used to model interactions among economic agents and
make predictions.
Applied in many other areas such as politics, law, biology, computer
science etc.
Game theory as a formal model:
Provides a precise language that can be used across varying contexts.
Allows us to test our intuition, and insights for logical consistency.
Clarifies the role of assumptions in arriving at particular predictions.

Some Questions in Analysing Strategic Interactions
Individuals within a group making interdependent choices:

What will each individual conjecture about the others’ choices?
What will an individual choose? (usually non-trivial because the best
choice depends on what others do)
What is the eventual outcome and could it have been improved for
the group as a whole?

Examples of Strategic Interactions
Oligopolies e.g. Coke-Pepsi etc.

Cartels e.g. OPEC, De Beers
Auctions e.g. Sotheby’s, treasuries.
R&D expenditures
Trench warfare in World War 1.
Animal conflicts over scarce resources.
Your team work for seminar presentations.

Elements of a Game
Who are the players?

What actions can they take?
What are the payoffs?
When does each player make her move?
What does each player know at each point in the game?
(Information structure)

Some Important Concepts
Complete Information: players know rules, payoffs, possible actions

of all players and this is common knowledge.
Perfect Information: at any point in time during the game players
know earlier moves by themselves and others and the outcome of all
preceding probabilistic events.
Extensive form: the time structure of the game (who moves when)
is accurately described in a game tree (captures dynamics)
Normal or Strategic form: summary of the game by describing all
possible “strategies” of a player (treat game as a simultaneous move
game)
In a simultaneous one-shot game, a strategy is just a choice or decision.
In sequential games, strategy is a complete contingent plan of
action (more on this later).

Noncooperative vs. Cooperative Game Theory
Noncooperative Games:
Unit of analysis is the individual player who acts to maximise her own
payoff subject to the rules of the game.
If the game outcome looks like “cooperation”, it is because it is in each
player’s unilateral interest to cooperate (we will see this emerge
sometimes in repeated games)
Will be the focus of the game theory part of this course.
Cooperative Games:
Unit of analysis is the group or the coalition.
Typically assumes unlimited communication, and unrestricted ability to
achieve agreements.
Focuses on the set of possible outcomes that can be achieved without
elaborating on the bargaining process itself.

Formal Definition of a Strategic-Form Game
Definition: A strategic form game is a triplet [N, (Si )i∈N , (ui )i∈N ] such
that:
N is a finite set of players, i = (1, 2, 3, ..., n);
Si is the set of strategies available to player i;
si ∈ Si is a strategy for player i;
ui : SQ→ R is the payoff (utility) function for player i where
S = i Si is the set of all strategy profiles.
Other useful notation:
s−i = [sj ]j̸=i denotes the vector of strategies of all players except i.
S−i = [S−j ]j̸=i denotes the set of strategies for all players except i.
(si , s−i ) ∈ S is a strategy profile or outcome.

Dominated Strategies
A Simple Example
B
L R
A U (1,1) (1,0)
D (0,1) (0,0)
Normal or Strategic form for 2 players

Row player (A) can choose U or D; column player (B) L or R.
Payoff pairs - first number for row player, second for the column player

A Simple Example
B
L R
A U (1,1) (1,0)
D (0,1) (0,0)
Normal or Strategic form for 2 players

Row player (A) can choose U or D; column player (B) L or R.
Payoff pairs - first number for row player, second for the column player
Rational players - A will never choose D and B will never choose R
(strictly dominated strategies)
This is not really a game - it is two independent decision problems!
Easy to predict (U, L) as outcome

The Prisoner’s Dilemma
B
Cooperate Defect
A Cooperate (4, 4) (0, 5)
Defect (5, 0) (1, 1)
Prisoner’s dilemma is a game because a player’s payoff is affected by

the other’s choice.

B
Cooperate Defect
A Cooperate (4, 4) (0, 5)
Defect (5, 0) (1, 1)
Prisoner’s dilemma is a game because a player’s payoff is affected by

the other’s choice.
Rational players will choose Defect (i.e. “confess” in the back-story).
Defect is a strictly dominant strategy – higher payoff than any
alternative strategy regardless of what the other players do.
“Cooperate” is a strictly dominated strategy – a rational player will
never choose such a strategy because there is a better alternative
regardless of what the other players do.

Both players defecting is the unique outcome.

It is suboptimal from the players’ perspective – Pareto inefficient
Both players cooperating is a Pareto improvement
But it cannot be achieved in this one shot game. Even if the players
talk to each other beforehand!
Popular game as it distills into a very simple form the tension between
group optimality (Pareto efficiency) and individual incentives – a
smple social dilemma.
Experimental evidence that some play dominated strategies even in
one-shot games:
Don’t understand the game?
Or not selfish?
Or think they are in a repeated game setting?

Formal Definition
Definition: A strategy si ∈ Si for player i is strictly dominated if there

exists some si′ ∈ Si such that:
ui (si′ , s−i ) > ui (si , s−i ) for all s−i ∈ S−i .
Definition: A strategy si ∈ Si for player i is weakly dominated if there

exists some si′ ∈ Si such that:
ui (si′ , s−i ) ≥ ui (si , s−i ) for all s−i ∈ S−i .

ui (si′ , s−i ) > ui (si , s−i ) for some s−i ∈ S−i .

Iterated Elimination of (Strictly) Dominated Strategies
B
L R
U (3, 6) (7, 1)
A M (5, 1) (8, 2)
D (6, 0) (6, 1)
A player may not have a (strictly) dominant strategy but still have
(strictly) dominated strategies.
In the game above, neither player has a strictly dominant strategy.

B
L R
U (3, 6) (7, 1)
A M (5, 1) (8, 2)
D (6, 0) (6, 1)
A player may not have a (strictly) dominant strategy but still have
(strictly) dominated strategies.
In the game above, neither player has a strictly dominant strategy.
But for player A, U is strictly dominated (by M).
Player B can assume that a (rational) player A will never play U.

B
L R
U (3, 6) (7, 1)
A M (5, 1) (8, 2)
D (6, 0) (6, 1)
We can delete U (as depicted).

In the remaining game, R is the dominant strategy for player B.
Realising that (rational) player B will always play R, player A will
choose M.
The predicted outcome is (M, R).

IEDS and Common Knowledge of Rationality
This procedure of iterated elimination of (strictly) dominated

strategies can be justified by use of common knowledge of
rationality:
Player A is rational i.e. won’t play a dominated strategy.
Player B knows player A is rational.
Player A knows that player B knows that player A is rational.
And so on...
Solving Prisoner’s dilemma only required that each player be rational
i.e. picks her dominant strategy. She does not have to assume
rationality of other players.
But in general, IEDS requires the assumption of common knowledge.

The Guessing Game
N participants are asked to guess a number between 0 and 100.

The winner is the number which is closest to 2/3rds of the mean of
the choices of all players.
Winner gains a fixed prize which is split if there is a tie.
What would you guess? And why?

The Guessing Game
N participants are asked to guess a number between 0 and 100.

The winner is the number which is closest to 2/3rds of the mean of
the choices of all players.
Winner gains a fixed prize which is split if there is a tie.
What would you guess? And why?
Game has been played with several types of participants:
Large and public experiments using newspapers (Financial Times,
Spektrum, Expansion)
Students (labs, take-home, class rooms)
Game theorists
See Bosch-Domenech et al. “One, two, (three), infinity,...: Newspaper
and lab beauty-contest experiments,” it American Economic Review 92
(5), 1687-1701.

The Guessing Game
Game can be solved by iterated elimination of dominated strategies:

2
1 A rational player will not choose numbers above 100 × 3 ≈ 66.67
Highest target number is
2 N × 100 2
× = 100 ×
3 N 3
Numbers higher than this are (weakly) dominated by 100 × 23 .

The Guessing Game

2
2 N × 100 2
× = 100 ×
3 N 3
2 A rational player would now not choose a number above
2
100 × 23 ≈ 44.44.
(Simultaneously) eliminate the possibility any rational player would
choose a number above 66.67.
Highest target number is now
2
2 N × 100 × 32 2
× = 100 ×
3 N 3

The Guessing Game

2
2 N × 100 2
× = 100 ×
3 N 3
2 A rational player would now not choose a number above
2
100 × 23 ≈ 44.44.
(Simultaneously) eliminate the possibility any rational player would
choose a number above 66.67.
Highest target number is now
2
2 N × 100 × 32 2
× = 100 ×
3 N 3

3 ...
The Guessing Game
All numbers are eliminated: only choosing 0 survives this (infinite)

iteration process
Requires common knowledge of rationality
You can also verify that all players choosing 0 is a Nash Equilibrium
(no unilateral incentive to deviate)

The Guessing Game – Experimental Results
Lab experiments with undergraduates

0 is rarely seen in initial play
Convergence to 0 with repetition

Also Known as the Beauty Contest
“...professional investment may be likened to those newspaper

competitions in which the competitors have to pick out the six prettiest
faces from a hundred photographs, the prize being awarded to the
competitor whose choice most nearly corresponds to the average
preferences of the competitors as a whole; so that each competitor has to
pick not those faces which he himself finds prettiest, but those which he
thinks likeliest to catch the fancy of the other competitors, all of whom are
looking at the problem from the same point of view. It is not a case of
choosing those which, to the best of one’s judgment, are really the
prettiest, nor even those which average opinion genuinely thinks the
prettiest. We have reached the third degree where we devote our
intelligences to anticipating what average opinion expects the average
opinion to be. And there are some, I believe, who practice the fourth, fifth
and higher degrees.”
Keynes (1936, p. 156)
Nash Equilibrium
A Motivating Example
H M L
H (0, 4) (4, 0) (5, 3)
M (4, 0) (0, 4) (5, 3)
L (3, 5) (3, 5) (6, 6)
No dominant strategies, no dominated strategies.

Nash Equilibrium
A Motivating Example
H M L
H (0, 4) (4, 0) (5, 3)
M (4, 0) (0, 4) (5, 3)
L (3, 5) (3, 5) (6, 6)
No dominant strategies, no dominated strategies.

Notice that when both agents play L neither gains from deviating
from L.
This is the spot where no player has a unilateral incentive to deviate
i.e. no profitable deviation conditional on the other player’s choice
(Nash Equilbrium)

Nash Equilibrium
Finding the Nash Equilibria – Best Responses
H M L
H (0, 4) (4, 0) (5, 3)
M (4, 0) (0, 4) (5, 3)
L (3, 5) (3, 5) (6, 6)
Find the best response of each player to all of the other’s strategies,
e.g.
If row plays H, then column’s best response is H.
Underline the corresponding payoff.

Nash Equilibrium
H M L
H (0, 4) (4, 0) (5, 3)
M (4, 0) (0, 4) (5, 3)
L (3, 5) (3, 5) (6, 6)
e.g.
If row plays M, then column’s best response is M.

Nash Equilibrium
H M L
H (0, 4) (4, 0) (5, 3)
M (4, 0) (0, 4) (5, 3)
L (3, 5) (3, 5) (6, 6)
e.g.
If row plays L, then column’s best response is L.

Nash Equilibrium
H M L
H (0, 4) (4, 0) (5, 3)
M (4, 0) (0, 4) (5, 3)
L (3, 5) (3, 5) (6, 6)
e.g.
If column plays H, then row’s best response is M.

Nash Equilibrium
H M L
H (0, 4) (4, 0) (5, 3)
M (4, 0) (0, 4) (5, 3)
L (3, 5) (3, 5) (6, 6)
e.g.
If column plays M, then row’s best response is H.

Nash Equilibrium
H M L
H (0, 4) (4, 0) (5, 3)
M (4, 0) (0, 4) (5, 3)
L (3, 5) (3, 5) (6, 6)
e.g.
If column plays L, then row’s best response is L.

Nash Equilibrium
H M L
H (0, 4) (4, 0) (5, 3)
M (4, 0) (0, 4) (5, 3)
L (3, 5) (3, 5) (6, 6)
e.g.
If column plays , then row’s best response is .
(There might be a tie in which case there is more than one strategy in
the best response. Underline all the payoffs corresponding to the best
response.)
The NE are the intersection of best responses - each player is playing
a best response to the others.

Nash Equilibrium
Formal Definition
Definition: A pure strategy Nash Equilibrium of a strategic game

[N, (Si )i∈N , (ui )i∈N ] is a strategy profile s∗ ∈ S such that for all i ∈ N:
ui (s∗i , s∗−i ) ≥ ui (si , s∗−i ) for all si ∈ Si .
No player has a unilateral incentive to deviate given the strategies of

other players.
Attractive notion - hard to argue that an outcome is an equilibrium if
one or more players wishes to deviate!

Nash Equilibrium
Should We Expect Outcomes to be Nash?
It is not straightforward to justify why we should expect Nash behaviour:

Introspection? Rational players should be able to reason their way
to Nash equilibrium i.e. realise that if play is not Nash, at least one
player (themselves and/or the others) would seek to deviate.
Self-enforcing? If rational players agree beforehand on a strategy
profile before independently choosing their actions, no player has an
incentive to deviate.
Learning/evolution? In repeated interactions, through trial and
error, non-Nash behaviour dies down.
BUT there are many Nash equilibria - how do we choose among
them?

Nash Equilibrium
Multiple Nash
B
Work Shirk
A Work (2, 2) (-1, 1)
Shirk (1, -1) (0, 0)
Partnership game; no dominant strategy and no dominated strategy

Work is best response to work, and shirk is best response to shirk
Two pure strategy Nash: (Work, Work) and (Shirk, Shirk)
Depending on your conjectures about your partner, good or bad
outcome possible
=⇒ possible to be caught in an “expectations trap”

Nash Equilibrium
Comparison with IEDS
Iterated elimination of (strictly) dominated strategies often have a

very imprecise prediction – sometimes no strategy is eliminated, so
there is no prediction.
If iterated elimination of strictly dominated strategies eliminate all but
a single combination of strategies (1 strategy remaining per player)
then that combination is the unique Nash equilibrium.
There is at least one Nash equilibrium in any finite game (finite
number of players and finite number of strategies).
Sometimes this Nash Equilibrium is probabilistic – in mixed
strategies.

Mixed Strategies
Matching Pennies Game
H T
H (1, -1) (-1, 1)
T (-1, 1) (1, -1)
Matching pennies - two players simultaneously show their coins.

Row would like to match while column would not.
This is a zero-sum game - in each cell, one player’s gain is the
other’s loss.

Mixed Strategies
H T
H (1, -1) (-1, 1)
T (-1, 1) (1, -1)
Matching pennies - two players simultaneously show their coins.

Row would like to match while column would not.
This is a zero-sum game - in each cell, one player’s gain is the
other’s loss.
No pure strategy Nash equilibrium.
How would you play if you were the row player?

Mixed Strategies
Mixed strategy: a probability distribution over two or more pure

strategies.
A mixed strategy Nash Equilibrium:
Players randomise over their pure strategies.
A player’s ‘mixing’ over her pure strategies leaves the other player(s)
indifferent between her pure strategies.
What is the mixed strategy NE in matching pennies?

Mixed Strategies
Keeping the Other Player “Guessing”
(q) (1−q)
H T
(p)
H (1, -1) (-1, 1)
(1−p)
T (-1, 1) (1, -1)
Suppose the row (player 1) plays H with probability p (and T with

probability 1 − p)
For what p is the column (player 2) indifferent between H and T?

Mixed Strategies
Keeping the Other Player “Guessing”
(q) (1−q)
H T
(p)
H (1, -1) (-1, 1)
(1−p)
T (-1, 1) (1, -1)
Suppose the row (player 1) plays H with probability p (and T with

probability 1 − p)
For what p is the column (player 2) indifferent between H and T?
If 2 chooses H, EU2 (H) = −1 × p + 1 × (1 − p) = 1 − 2p
If 2 chooses T, EU2 (T ) = 1 × p − 1 × (1 − p) = 2p − 1
To make player 2 indifferent between H and T, player 1 must set p
such that EU2 (H) = 1 − 2p = 2p − 1 = EU2 (T ) ⇒ p ∗ = 0.5
q ∗ = 0.5 makes player 1 indifferent between H and T.
Mixed NE: [(0.5, 0.5), (0.5, 0.5)]

Mixed Strategies
Best Responses in Matching Pennies
Best response of player 2 to any p by player 1?

Mixed Strategies
Best Responses in Matching Pennies
Best response of player 2 to any p by player 1?

For p < 0.5, EU2 (H) = 1 − 2p > 2p − 1 = EU2 (T ) ⇒ q ∗ (p) = 1
For p > 0.5, EU2 (H) = 1 − 2p < 2p − 1 = EU2 (T ) ⇒ q ∗ (p) = 0;
For p = 0.5, EU2 (H) = EU2 (T ) ⇒ q ∗ (p) ∈ [0, 1]
We can graph the best response correspondences for the two players:

Mixed Strategies
Formal Definition
Notation:
∆i be the set of probability measures over the pure strategy set Si
σi ∈ ∆i isQa mixed strategy of player i
σ ∈ ∆ = i∈N ∆i is a mixed strategy profile for the game.
σ−i ∈ ∆−i is a mixed strategy profile for players other than i
Definition: A mixed strategy profile σ ∗ is a Nash equilibrium if for
each player i:
∗ ) ≥ u (σ , σ ∗ ) for all σ ∈ ∆
ui (σi∗ , σ−i i i −i i i
It suffices to check only for profitable pure strategy deviations when
checking if a given mixed strategy profile is Nash; i.e.
“for all si ∈ Si ” rather than “for all σi ∈ ∆i ”.

MiA T2 StaticGames

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

MiA T2 StaticGames

Uploaded by

Copyright:

Available Formats

Microeconomic Analysis (806L1)

Topic 2: Game Theory

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 1 / 31

“No man is an island.”

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 2 / 31

Individuals within a group making interdependent choices:

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 3 / 31

Oligopolies e.g. Coke-Pepsi etc.

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 4 / 31

Who are the players?

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 5 / 31

Complete Information: players know rules, payoffs, possible actions

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 6 / 31

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 7 / 31

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 8 / 31

Normal or Strategic form for 2 players

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 9 / 31

Normal or Strategic form for 2 players

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 9 / 31

Prisoner’s dilemma is a game because a player’s payoff is affected by

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 10 / 31

Prisoner’s dilemma is a game because a player’s payoff is affected by

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 10 / 31

Both players defecting is the unique outcome.

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 11 / 31

Definition: A strategy si ∈ Si for player i is strictly dominated if there

ui (si′ , s−i ) > ui (si , s−i ) for all s−i ∈ S−i .

Definition: A strategy si ∈ Si for player i is weakly dominated if there

ui (si′ , s−i ) ≥ ui (si , s−i ) for all s−i ∈ S−i .

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 12 / 31

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 13 / 31

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 13 / 31

We can delete U (as depicted).

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 14 / 31

This procedure of iterated elimination of (strictly) dominated

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 15 / 31

N participants are asked to guess a number between 0 and 100.

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 16 / 31

N participants are asked to guess a number between 0 and 100.

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 16 / 31

Game can be solved by iterated elimination of dominated strategies:

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 17 / 31

Game can be solved by iterated elimination of dominated strategies:

Numbers higher than this are (weakly) dominated by 100 × 23 .

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 17 / 31

Game can be solved by iterated elimination of dominated strategies:

Numbers higher than this are (weakly) dominated by 100 × 23 .

All numbers are eliminated: only choosing 0 survives this (infinite)

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 18 / 31

Lab experiments with undergraduates

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 19 / 31

“...professional investment may be likened to those newspaper

No dominant strategies, no dominated strategies.

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 21 / 31

No dominant strategies, no dominated strategies.

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 21 / 31

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 22 / 31

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 22 / 31

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 22 / 31

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 22 / 31

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 22 / 31

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 22 / 31

Matthew Embrey (University of Sussex) Microeconomic Analysis (806L1) Autumn 2023 22 / 31