Department of Economics A non-strategic

ESSEC Business School

Gorkem Celik
(Decision Theoretical)
Problem
ECOA31303 • The family of a good friend of yours is
NOBEL IDEAS IN ECONOMICS running a winery in the Côte Rôtie region.
She offers you an investment opportunity.
• You will decide on how much money to
Game Theory
invest on her winery.
John Nash (Nobel Prize in 1994) • If this will be a good year for Côte Rôtie,
you double your money.
Modeling strategic interactions
• If it is a bad year, you lose half your money.
1 2

A non-strategic (Decision Theoretical) An example to a strategic setting:

Problem Tragedy of the Commons
• Whether it will be a good year or a bad year is
completely out of our control. • Let’s play an in-class game to get a feeling of the
– determined by environmental factors. decision-making process in a strategic setting.
• Let’s say that the probability that this will be a good • You will decide how many cows to buy: s
year is ½ . • The cost of buying a cow is 500 euros.
• How much money you will invest depends on your risk • The yield (value) from each cow depends on the
total number of cows bought by all the villagers:
preferences
– whether you are risk averse or risk lover, etc. 1000 − ෍ 𝑠𝑗
• more specifically on your cash utility function that you
may remember from your microeconomics courses. • Choose the number of cows you would like to buy
between 0 and 10.
3 [Link]
 Players
Game Theory • Each decision-maker in a game is called a
player.
• All games have three elements: – can be an individual, a firm, an entire nation.
– players • Each player has the ability to choose among
– strategies a set of possible actions.
– payoffs • The specific identity of the players is
irrelevant.
• For the commons game, the set of players N
is the set of students enrolled in this class
(say, 40 students).
5 6

Payoffs
Strategies • The final returns to the players at the end of
the game are called payoffs.
• Each course of action open to a player is • Payoff of a player depends on his strategy
called a strategy. as well as the strategies of the other players.
• A complete contingent plan • Payoffs are usually measured in terms of
• Strategies can be very simple or very utility.
complex, – monetary payoffs are also used.
– can be in the form of contingent strategies. • We assume that players can rank the
• In the commons game, the set of strategies payoffs associated with a game.
for each player i is Si = [0,10].
7 8
 Payoffs
• What is the payoff of player i in the commons The Prisoners’ Dilemma
game? The most famous game of all times:
• As we have just seen, payoff is not necessarily
• Two criminals named 1 and 2 have been
identical to the monetary gains. apprehended by the police and are being questioned
• But if we are comfortable with that separately.
• They are jointly guilty of the crime. The police has
assumption, some inconclusive evidence against the criminals.
Each criminal can choose either to confess or deny.
𝑣𝑖 𝑠𝑖, 𝑠 − 𝑖 = 1000 − ෍ 𝑠𝑗 − 500 𝑠𝑖
𝑗∈𝑁
• If both confess, each gets a 4 year jail sentence.
𝑠𝑖 ∈ Si = [0,10]: player i’s own strategy. • If both deny, each gets a 2 year sentence.
𝑠 − 𝑖 : the strategies of the other players (rivals). • If one confesses and the other denies, the confessor
9 serves 1 year and the denier serves 5 years in jail. 10

Normal Form Representation of Normal Form Representation of

The Prisoners’ Dilemma The Prisoners’ Dilemma
Player 2’s Strategies
Deny Confess The Payoff matrix
2’s Strategies
1: 2 years 1: 5 year Deny Confess
Deny
Player 1’s 2: 2 years 2: 1 years
Strategies 1: 1 years 1: 4 years Deny -2 , -2 -5 , -1
Confess
2: 5 year 2: 4 years 1’s Strategies
Confess -1 , -5 -4 , -4
Assume that each criminal is trying to minimize his
jail term. 11 12
 The Prisoners’ Dilemma Strict Domination
• First, Row player (Player 1) has a strict advantage
to confessing, no matter what Column player (Ply 2) For player i, strategy 𝑠𝑖′′ strictly dominates 𝑠𝑖′ if
is going to do.
𝑣𝑖 𝑠𝑖′′, 𝑠 − 𝑖 >𝑣𝑖 𝑠𝑖′, 𝑠 − 𝑖
• If Column confesses, Row gets –4 for confessing,
–5 for denying, and thus is better off confessing. for every list 𝑠 − 𝑖 of other players’ strategies.
• Similarly, if Column denies, Row gets –1 for • Strategy 𝑠𝑖 ′ is strictly dominated by strategy s𝑖′′ .
confessing, –2 for denying, and is better off
confessing. • Strategy 𝑠𝑖′′ is strictly dominant if it strictly
• Either way, no matter what Column does, Row dominates all the other strategies of player i.
should choose to confess. • In prisoners’ dilemma, for both players, strategy
• Symmetric game: The same argument applies to Confess strictly dominates strategy Deny.
Column player too.
• Therefore, our prediction for this game is that both13
Confess is the strictly dominant strategy.
14
players will confess.

The Prisoners’ Dilemma The Prisoners’ Dilemma: Efficiency

Prediction • According to Wilfredo Pareto, an outcome is efficient for
the society if there is no other alternative outcome which
• A strategy profile 𝑠𝐷 𝐷, 𝑠 𝐷
= (𝑠1 2 … 𝑠𝑛 is a 𝐷)
makes someone better off without hurting someone else.
Strictly Dominant Strategy Equilibrium if 𝑠𝑖 𝐷 ∈ 2’s Strategies
Deny Confess
𝑆𝑖 is a strictly dominant strategy for all 𝑖 ∈ 𝑁.
Deny -2 , -2 -5 , -1

• (Confess, Confess) is a strictly dominant strategy 1’s Strategies

Confess -1 , -5 -4 , -4
equilibrium for prisoners’ dilemma. • (Confess, Confess) is not Pareto efficient.
• It is our prediction for this game. – An ironclad agreement by both prisoners to deny will give them
the lowest amount of joint jail time.
• Because we conjectured that no player will play a
– Both players would be better off if they could play (Deny, Deny).
strictly dominated strategy.
• But (Deny, Deny) is not stable.
• In other words, we assumed that each player is – Either player would individually prefer to deviate and choose
15 16
rational. Confess.
 Prisoners’ Dilemma
The Prisoners’ Dilemma: Efficiency
• Fits to a wide variety of real-life strategic
• Contrast this with the efficiency of competitive interactions.
markets result in microeconomics. • Example 1: You are sharing a dormitory room.
• There is no centralized planner, nevertheless, the • You prefer a tidy room to a messy one.
output traded in a competitive market is the social • You would rather that your roommate cleans up
surplus maximizing one.
the room instead of you.
• « It is as if an ‘Invisible Hand’ guided the competitive
market to the efficient level of production and • And you would absolutely not like to clean up
consumption, » your roommate’s mess.
Adam Smith, 1776, An Inquiry in the Nature and Causes of the Wealth
of Nations.

17 18

The Roommates’ Dilemma Other Dilemmas from Business

2’s Strategies World and International Relations
Clean up Mess up
• Pricing your products – at the same time as
Clean
up
2,2 0,3 your competitors.
1’s Strategies
Mess
3,0 1,1
• Choosing your advertisement budget –
up
relatively higher budget implies higher
• If these are the payoffs, for both players, market share.
strategy Mess up strictly dominates strategy • Defense spending by the members of the
Clean up. same military alliance.
• Then, our prediction is that the room will • Breaking an armistice or not.
remain messy.
• The tragedy of commons.
• Is this Pareto efficient? 19 20
 Tragedy of Commons
• For our in-class game, recall that for each player i, Tragedy of Commons
– Cost of a cow: 500
– Return from a cow: 1000 − σ𝑗∈𝑁 𝑠𝑗 • Each villager decides how many cows to keep.
• Return will be higher than the cost, even if all 40 • The villager receives the benefits from an
additional cow.
villagers chose the maximum number of cows 10.
• But the damage to the common is shared by the
• 𝑠𝑖 =10 is a strictly dominant strategy in Si = [0,10]. entire group.
• Equilibrium payoff: – Overgrazing and destruction of the commons.
𝑣𝑖 = 1000 − 400 − 500 10 = 1000
• Is this a Pareto Efficient outcome?
• Imagine that the villagers could agree to own only
5 cows each…
21 22

Tragedy of the Commons

Similar Examples
A 3x3 Game
• Fisheries / Oceans • Not all games have strictly dominant strategies.
• Use of Water resources • But we can still use strict dominance to make
predictions.
• Global warming
– Most governments agree that we have to reduce carbon • Consider the following capacity-setting game.
emissions. • Firms 1 and 2 will choose their production capacity
– There is a private political cost to emission reductions for a new product.
(Trump’s election in the US, the “gilets jaunes” in France),
• Market share of each firm depends on its individual
but the benefits are enjoyed by all.
capacity choice.
– Securing compliance is an important part of international
environmental treaties. • Market price will depend on the total capacity.
– Paris agreement (COP 21 in 2015) was to enter into force • Similar to the Cournot competition game.
when joined by at least 55 countries representing 55% of
24
emissions (satisfied by the end of 2016). Why?
 A 3x3 Game A 3x3 Game
• Suppose that industry profits are maximized when • Notice that there is no strictly dominant strategy.
both firms choose low capacity (higher price).
• Any strictly dominated strategy for Firm 1?
• But Firm 2 has an individual incentive to choose a
little higher capacity than its competitor. • What about Firm 2?
Firm 2’s Strategies Firm 2’s Strategies
High Medium Low High Medium Low

High 3,3 3,2 2,1 High 3,3 3,2 2,1

Firm 1’s Firm 1’s
Medium 2,5 4,4 3,3 Medium 2,5 4,4 3,3
Strategies Strategies
Low 1,4 3,6 5,5 Low 1,4 3,6 5,5
25 26

Another 3x3 Game Nash Equilibrium in Games

• But there are other games where there is no
dominated strategy for any player! • In a Nash equilibrium, each player is choosing
her best strategy, given the strategy of the
Player 2’s Strategies other players.
Left Center Right • Each player is playing her best response to
the play of the other players.
Top 0,7 2,5 7,0
• A strategy profile 𝑠 ∗ = (𝑠1∗ , 𝑠2∗ , … 𝑠𝑛 ∗ ) is a
Player 1’s
Middle 5,2 3,3 5,2 Nash Equilibrium if
Strategies
Bottom 7,0 2,5 0,7 𝑣𝑖 𝑠𝑖 ∗ , 𝑠 − 𝑖 ∗ ≥ 𝑣𝑖 𝑠𝑖, 𝑠 − 𝑖 ∗
for each strategy 𝑠𝑖 ∈ 𝑆i of each player i ∈ N.
27 28
 Nash Equilibrium in Games Finding Nash equilibrium in two
In a Nash Equilibrium,
player games
• Each player chooses his best strategy given his
Player 2’s Strategies
beliefs about the other players’ strategies.
Left Center Right
• Moreover, each player’s beliefs are correct:
They are consistent with what the other players Top 0,7 2,5 7,0
choose. Player 1’s
Middle 5,2 3,3 5,2
• This means that, at Nash Equilibrium, no player Strategies
has an incentive to change his behavior, after Bottom 7,0 2,5 0,7
learning how the others are behaving.
• In this game, strategy profile (Middle, Center)
• No player has a profitable deviation. is a Nash equilibrium.
29 30

Electoral Competition Electoral Competition

• Originally due to Hotelling (1929), a model of
• The Game: 2 candidates choosing their
product differentiation.
policy positions.
• Applied to electoral competition, Downs (1957).
• The candidate with the highest number of
• There is a large number of voters each with an votes wins, the other one loses.
ideal policy.
• If the vote shares are the same, they tie.
– The ideal income tax level: 35%? 45%? 70%?
• Candidates care only about the outcome of
• The voters prefer policies that are closer to their
the election: they prefer to win rather than
ideal policy.
tie, and prefer to tie rather than lose.
• The distribution of voters’ ideal points is common
knowledge. Median of this distribution is m. 31 32
 Electoral Competition Nash Equilibrium
Nash Equilibrium
• Augustin Cournot (1838)
• Both candidates choosing the median policy • Joseph Bertrand (1883)
position m is a Nash equilibrium. • Harold Hotelling (1929)
– The candidates tie in equilibrium. If a candidate • John Nash (1950, in his PhD Thesis)
deviates, she loses.
Why do we call it the Nash Equilibrium today?
• There is no other Nash equilibrium.
Because Nash is the one who formalized this way
– If your rival is not choosing m, you can always of reasoning as an equilibrium concept that can
guarantee winning by deviating to m. be used for different social interactions.
• This is known as the Median Voter Theorem.
33 34

Nash Equilibrium
John Nash (1928-2015). Nobel prize in 1994. Nash Equilibrium in Games
Life of John Nash has been the subject of the book Why is Nash Equilibrium thought to be a good
A Beautiful Mind, by Sylvia Nassar (1998) prediction in a game?
and a movie by the same name (2001) • Rationality or common knowledge of Rationality
are not sufficient.
• Nash Equilibrium as a Self-enforcing agreement.
• Stable Social Conventions
– If one of the players reveals the equilibrium strategy
he will use, the other player cannot benefit.
– This is not the case with non-equilibrium strategies.
• Evolution
35
• Learning [Link]
36
 Other Examples:
Other Examples to Nash Equilibria « The bar scene » from the movie
The golden balls game [Link]
d_dtTZQyUM
• There are 5 friends.
• Each decides whether to pursue the
only blonde in the bar, or target one
of the many brunettes.
• Blonde: payoff 2 only in the case
that there is a unique pursuer.
Otherwise payoff zero.
Split or Steal:
• Brunette: the safe option – payoff 1.
[Link]
37 38

Game Theory: Historical Perspective Game Theory: Other big contributors

Theory of Games and Economic Behavior (1944), book Nobel Prizes in Economics:
by John von Neumann and Oskar Morgenstern. 1994 - John Nash, Reinhard Selten (Dynamic Games),
• Normal-form representation and John Harsanyi (Bayesian Games).
– players, strategies, payoffs. 1996 - James Mirrlees and William Vickrey: Incentives.
2001 - George Akerlof (lemons), Michael Spence
Limitations: (signaling), Joseph Stiglitz (screening): Asymmetric info.
• two-player, zero-sum games (matching pennies). 2005 - Robert Aumann (repeated games) and Thomas
Schelling (multiple equilibria).
• in the aftermath of the world wars.
2007 - Leonid Hurwicz, Eric Maskin, Roger Myerson:
Generalization to n-player games: cooperative game mechanism design.
theory approach – coalition of players coordinating their
strategies. 2014 - Jean Tirole: Market power and Regulation.
John Nash (1950s): non-cooperative game theory – 2016 - Oliver Hart and Bengt Holmström: Contracts.
individual decision making for non-zero-sum, n players. 2020 - Robert B. Wilson and Paul R. Milgrom: Auctions.
 Game Theory beyond Nash Game Theory beyond Nash
Sequential Games
Bayesian Games
• A committee of three members must choose
• Two players. Each has an envelope with 5, 10,
one of three options: A, B, C.
20, 40, 80, or 160 euros.
• Preference rankings:
– Player I: AÂBÂC
• One of the envelopes has twice as much money.
– Player II: BÂCÂA • Each player sees what is in the envelope and
– Player III: CÂAÂB then decides whether they want to exchange.
– They vote with majority voting. • If both agree to exchange, then they exchange.
• In case of a tie, Player III decides. • Suppose you see 20 euros in your envelope.
Should you exchange?
What do you think will happen? 41 42