Game Theory: Guillem Roig

Game Theory
Introduction
Guillem Roig
Universidad del Rosario
1 / 28
1.1 What is Game Theory?
2 / 28
What is Game Theory ?
I In your previous economics classes, you have studied decision making

by individuals (consumers/firms) in environments in which the
decision of one individual does not affect (is not affected by) the
decisions of another individual.
I However, in many situations the payoff of an individual depends not

only on his own decisions, but also on the decisions made by others.
There is Strategic Interdependence.
3 / 28
What is Game Theory?
I Strategic interdependence implies that:

I Each agent’s optimal behavior depends on the behavior of the other
agents’ behavior directly.
I Examples of such situations abound in real life:
I Survival of species.
I Oligopolistic competition.
I A tennis match.
I The tool we use to analyze situations with strategic interdependence is

Non-Cooperative Game Theory
4 / 28
What is Game Theory?
Game + Theory
1. Game: like in common games, individuals (players) are in the search of

the best strategies (that might depend on what the others do) in order to
win (maximize payoff).
2. Theory: We will describe strategic situations in an abstract way, and

propose solutions concepts to attempt to predict behavior by players.
5 / 28
Key Assumptions in Game Theory
Two Assumptions
I Decision-makers (agents, individuals, players, etc.) pursue well-defined

objectives.
=⇒ they are RATIONAL.
I They take into account their knowledge or expectations of other

decision-makers’ behavior.
=⇒ they REASON STRATEGICALLY.
6 / 28
The Objective of Game Theory
What are we after?
I We want to reach a prediction of the game. A solution is a systematic

description of outcomes that may emerge in a game.
I This is important if we need to predict the behavior of people, firms

etc... (design of taxes in a country, incentives for workers, market design
where firms compete...)
7 / 28
1.2 Examples of Strategic Situations
8 / 28
Examples of Strategic Situations
Hide and Seek
I Camilo and Marta are playing hide and seek. Camilo it is.
I Marta decides where to hide among a set of (let’s say two) hiding
places: back of the house (H) or behind the trees (T).
I Camilo chooses one place to seek either (H) or (T).
I If Camilo finds Marta (he wins and she loses), if Camilo does not find
Marta (he loses and she wins).
9 / 28
Hide and Seek
I To describe a game we need to specify:
1. The players: Who is involved?

Marta and Camilo.
2. The rules: Who moves when? What are the possible actions? What do they
know when they move?
Marta moves first, decides where to hide. Camilo moves second, he doesn’t
know where Marta is hidden and needs to decide where to seek.
3. The outcomes: For each possible set of actions what happens?
If Camilo finds Marta, he wins and she loses. If Camilo doesn’t find Marta,
he loses and she wins.
4. The Payoffs: Players’ preferences over outcomes.
Both prefer to win.
10 / 28
Subscription to The Economist
I What are the benefits of reading The Economist?
I Consider first a situation without strategic interaction.

I Strategic Situation: Reading The Economist gives a competitive edge with respect
to my peers. Consider a class with two students.
I Nobody reads (status quo payoff of 0).
I One student reads and the other does not. The party who reads gets the
competitive edge (payoff of +1). The party who does not read is in a
competitive disadvantage (payoff of −2).
I Both read. There is no competitive edge and both students pay the
subscription fee (payoff of −1).
Question: what are the players, actions and payoffs?
11 / 28
Pirate Game (Moulin, 1988)
I 10 pirates have found a treasure with 100 gold coins.
I The problem that they confront is how to share this treasure.
I The mechanism implemented is that the oldest pirate (pirate 10) makes a
proposal, which is voted.
I If at least half of the pirates approves the proposal (#Yes ≥ 5 and who
makes the proposal also votes) the proposal is implemented.

I Otherwise (#Yes < 5) the oldest pirate, who made the proposal, is thrown
to the sea where there are many hungry sharks.

I Then the game continues and the second oldest pirate (pirate 9)
makes the proposal. The oldest pirate is not around anymore and
there are only 9 pirates left.
I ···
I All pirates on board are rational, there is no rum left, and want first to live.
Question: what are the players, actions and payoffs?
12 / 28
We are not yet adding formality into the model.
I The players: 10 pirates.

I The strategies: Depending on the role of the pirate:
I The pirate that makes a proposal offers an allocation of the treasure
specifying the number of gold coins to each pirate.
I The other pirates vote, (yes or no) to the allocation of gold coins.
I The payoffs: The number of coins specified in the allocation that is
accepted by simple majority.
13 / 28
I Any suggestion of the prediction of the game?

I Which one of these allocations do you find more plausible? (the
elements in the vector are ordered by age of the pirate, from oldest to
youngest).
I soldest = (10, 10, 10, 10, 10, 10, 10, 10, 10, 10).
I soldest = (20, 0, 20, 0, 20, 0, 20, 0, 20, 0).
I soldest = (96, 0, 1, 0, 1, 0, 1, 0, 1, 0).
I We will later see that we solve these type of games by the procedure of
backward induction.
14 / 28
1.3 Decision Theory (one agent)
15 / 28
Decision Theory
In the theory of rational choice, a decision-maker chooses the best action
according to her preferences, among all the actions available to her.
I A: set of actions available.

I X: set of outcomes.
I L(X): set of probability distributions (lotteries) on X.
I A probability
P distribution on X is a mapping p : X → [0, 1] such
that x∈X p(x) = 1.
I u : X → R: utility function of an agent over sure outcomes.
I g : A → L(X): outcome function that maps each possible action taken
by the agent into a probability distribution over the set of outcomes.
16 / 28
Decision Theory
I As to the agent’s preferences on the lottery L(X) we assume that

satisfy the properties guaranteeing that there exists a utility function
U : L(X) → R that represent preferences and satisfies the expected
utility property, requiring:
I for all p, p0 ∈ L(X), p p0 if and only if U(p) ≥ U(p0 ), and

I for all p ∈ L(X)
X
U(p) = p(x) · u(x).
x∈X
I Let H : A → R be the induced utility function on the available set of

actions, namely for each a ∈ A,
H(a) = U(g(a)).
17 / 28
Decision Theory
One-agent decision problem
I Choose an action a in order to

max H(a)
s.t. a ∈ A.
I The solution to this constrained maximization problem is represented

by a∗ .
18 / 28
An Example of Decision Theory
To put all those elements into context, consider that the best player in football
history (Messi) is kicking a penalty. Note that in a decision problem there is
no goal keeper.
I Set of actions, A : {Left, Middle, Right}.

I Outcomes, X : {Goal, No Goal}.
I Utility over outcomes, u : X → R. If (Goal) Messi gets 1 if (No Goal)
Messi gets −1.
I Actions mapping to lotteries over outcomes g : A → L(X).
I Left: p = 0.8 (Goal), and p = 0.2 (No Goal).
I Middle: p = 0.99 (Goal), and p = 0.01 (No Goal).
I Right: p = 0.7 (Goal), and p = 0.3 (No Goal)l.
19 / 28
An Example of Decision Theory
I The induced utility function H : A → R.

I H(Left) = 0.8 × 1 + 0.2 × (−1) = 0.6.
I H(Middle) = 0.99 × 1 + 0.01 × (−1) = 0.98.
I H(Right) = 0.7 × 1 + 0.3 × (−1) = 0.4.
I What action does Messi choose?

max H(a)
s.t. a ∈ A.
The solution of the Messi’s decision problem is a∗ = Middle
20 / 28
1.4 Decision Theory (two players): Game
Theory
21 / 28
Game Theory
I A game is a formal representation of a situation when there are more

than one decision-makers/players and each player’s payoff possibly
depends on the actions taken by the other players.
I A game is a strategic situation characterized by four components:
1. A set of people involved called players.

2. A set of possible actions available to each player.
3. Rules determining the order in which players make their moves.
4. A description of how each player feels about the possible outcomes
of the game described by a payoff or utility function.
22 / 28
Game Theory
I Set of players (agents) I = {1, 2}.
I Let A1 be the set of actions of player 1.
I Let A2 be the set of actions of player 2.
I A pair (a1 , a2 ) ∈ A1 × A2 is a profile of actions.
I For each player i = 1, 2,
I Let ui : X → R be the agent i’s utility function on sure outcomes.

I Ui : L(X) → R be agent i’s utility function on probability
distributions on X.
I Let g : A1 × A2 → L(X), be the outcome function that maps each

possible profile of actions into a probability distribution over X.
23 / 28
Game Theory
I Let Hi : A1 × A2 → R be the induced utility function on the set of

profiles of actions. Namely for each (a1 , a2 ) ∈ A1 × A2 ,
Hi (a1 , a2 ) = Ui (g(a1 , a2 )).
I A player may not know g(a1 , a2 ) if he does not know the other player’s
action, perhaps because they are taken simultaneously.
I Denote by ae2 to be the action that player 1 expects player 2 will
choose. Similarly, ae1 denotes the action that player 2 expects player
1 to take.
24 / 28
Game Theory
Best Responses
I Given action ae2 , player 1 chooses action a1 in order to:
max H1 (a1 , ae2 )

s.t. a1 ∈ A1 .
I The solution is represented by a1 (ae2 ) (Best response).
I Given ae1 , player 2 chooses a2 in order to:
max H2 (ae1 , a2 )
s.t. a2 ∈ A2 .
I The solution is represented by a2 (ae1 ) (Best response).
25 / 28
Game Theory
Equilibrium
I Best response, given the beliefs: a∗1 = a1 (ae2 ) and a∗2 = a2 (ae1 ).
I Beliefs are correct ae1 = a∗1 and ae2 = a∗2 .
I An equilibrium of the game is (a∗1 , a∗2 ) ∈ A1 × A2 such that

a∗1 = a1 (a∗2 ) and a∗2 = a2 (a∗1 ).
I We will later say that the intersection of best responses forms the
equilibrium(a) of the game.
26 / 28
1.5 Non-cooperative versus Cooperative Games
27 / 28
Non-cooperative versus Cooperative Games
1. Non-Cooperative Game Theory: The unit of analysis is an individual.
I In this games, every player looks after its self-interest.

I Positive approach: equilibrium analysis. In an equilibrium, an
outcome must have the property that no player, by unilaterally
changing his behavior, induces a better outcome.
2. Cooperative Game Theory: The unit of analysis are groups and

subgroups of individuals that can play joint (cooperative) actions.
I The assumption is that there is no constraint on the communication

among players.
I Interest on the type of agreements that players can reach.
28 / 28

Game Theory: Guillem Roig

Uploaded by

Copyright:

Available Formats

Game Theory: Guillem Roig

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

Game Theory: Guillem Roig

Uploaded by

Copyright:

Available Formats

Game Theory

Universidad del Rosario

I In your previous economics classes, you have studied decision making

I However, in many situations the payoff of an individual depends not

I Strategic interdependence implies that:

I The tool we use to analyze situations with strategic interdependence is

1. Game: like in common games, individuals (players) are in the search of

2. Theory: We will describe strategic situations in an abstract way, and

I Decision-makers (agents, individuals, players, etc.) pursue well-defined

=⇒ they are RATIONAL.

I They take into account their knowledge or expectations of other

=⇒ they REASON STRATEGICALLY.

What are we after?

I We want to reach a prediction of the game. A solution is a systematic

I This is important if we need to predict the behavior of people, firms

Hide and Seek

I Camilo chooses one place to seek either (H) or (T).

1. The players: Who is involved?

I Consider first a situation without strategic interaction.

Question: what are the players, actions and payoffs?

makes the proposal also votes) the proposal is implemented.

to the sea where there are many hungry sharks.

Question: what are the players, actions and payoffs?

We are not yet adding formality into the model.

I The players: 10 pirates.

I Any suggestion of the prediction of the game?

I A: set of actions available.

I As to the agent’s preferences  on the lottery L(X) we assume that

I for all p, p0 ∈ L(X), p  p0 if and only if U(p) ≥ U(p0 ), and

I Let H : A → R be the induced utility function on the available set of

One-agent decision problem

I Choose an action a in order to

I The solution to this constrained maximization problem is represented

I Set of actions, A : {Left, Middle, Right}.

I The induced utility function H : A → R.

I What action does Messi choose?

The solution of the Messi’s decision problem is a∗ = Middle

I A game is a formal representation of a situation when there are more

I A game is a strategic situation characterized by four components:

1. A set of people involved called players.

I Set of players (agents) I = {1, 2}.

I Let A1 be the set of actions of player 1.

I Let A2 be the set of actions of player 2.

I A pair (a1 , a2 ) ∈ A1 × A2 is a profile of actions.

I For each player i = 1, 2,

I Let ui : X → R be the agent i’s utility function on sure outcomes.

I Let g : A1 × A2 → L(X), be the outcome function that maps each

I Let Hi : A1 × A2 → R be the induced utility function on the set of

Hi (a1 , a2 ) = Ui (g(a1 , a2 )).

I Given action ae2 , player 1 chooses action a1 in order to:

max H1 (a1 , ae2 )

I The solution is represented by a1 (ae2 ) (Best response).

I Given ae1 , player 2 chooses a2 in order to:

I The solution is represented by a2 (ae1 ) (Best response).

I Beliefs are correct ae1 = a∗1 and ae2 = a∗2 .

I An equilibrium of the game is (a∗1 , a∗2 ) ∈ A1 × A2 such that

1. Non-Cooperative Game Theory: The unit of analysis is an individual.

I In this games, every player looks after its self-interest.

2. Cooperative Game Theory: The unit of analysis are groups and

I As to the agent’s preferences on the lottery L(X) we assume that

I for all p, p0 ∈ L(X), p p0 if and only if U(p) ≥ U(p0 ), and