Game Theory

Lecture 1 Basics

Young Joon Park

HSBC Business School

Peking University

February/2023

Young Joon Park (PHBS) Game Theory February/2023 1 / 31

Some Motivational Stories for Game Theory
Fight against Terrorism

CAAC (Civil Aviation Administration of China) conducted thorough

historical research on terrorism attacks in the past few decades and
reported that males with age between 30 and 55 committed 80% of
such attacks (including attempted attacks).
Following this investigation, CAAC adjusted the field manual to
mainly focus on fortifying search practice for male passengers with
age between 30 and 55.

Young Joon Park (PHBS) Game Theory February/2023 2 / 31

Some Motivational Stories for Game Theory
Long-run Interationcs

Dianping had enjoyed their position as the dominating shared-bike

provider in Shenzhen area. Once they learned Didi is considering
moving into Shenzhen market, Dianping announced they will double
their bicycle unit in greater Shenzhen area.
This move is costly for Dianping since this exceeds their optimal
number of bicycles based on Shenzhen’s market size.
How can this decision by Dianping be justified? What can be Didi’s
reaction?

Young Joon Park (PHBS) Game Theory February/2023 3 / 31

Some Motivational Stories for Game Theory
Issues with Demanding Salary

Mr. Wu runs a small business and recently received a verbal o↵er

from one of his business partners to join them as a local managing
officer. His local business is not growing, and the o↵er comes from a
much larger and fast-growing company. Naturally, he is seriously
considering the o↵er.
He is struggling, however, on what his asking salary should be. He
does not have much knowledge regarding the ongoing rate.
What kind of information does he need to make the best decision?

Young Joon Park (PHBS) Game Theory February/2023 4 / 31

Some Motivational Stories for Game Theory
Kaynes’s 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.”
-John Maynard Keynes, The General Theory of Employment, Interest, and
Money
Young Joon Park (PHBS) Game Theory February/2023 5 / 31
Game Theory for International Relations

Game theory is important for international relations theory because it

provides a useful way to analyze and understand the behavior of
decision makers related to international relations (di↵erent nations,
economics agents in international context, etc) in the international
system.
Game theory provides a framework for understanding how decision
makers interact and how they make decisions based on their goals and
interests and how they react to each others’ actions.
Game theory can also be used to develop strategies and anticipate
how opponents may respond, allowing countries to better prepare for
potential conflicts.

Young Joon Park (PHBS) Game Theory February/2023 6 / 31

Review of Economics and Mathematics

In order to improve the understanding of Game Theory, basic

knowledge of Economics and Mathematics is vital
We will limit our coverage to basic topics and key ideas

Young Joon Park (PHBS) Game Theory February/2023 7 / 31

DM’s problem

Economics studies interaction of individual decision makers.

Particularly we are interested in decision makers (DM)’ choice
behavior
DM must choose one alternative x from a choice set X .
Example of choice sets
I {Game Theory, Strategic Management, Corporate Finance,
Communication Theory}
I {Walk, Take a taxi, Take a bus, Ride a bike}
I {Start a war, Ask the UN for arbitration, O↵er a peace treaty}
I {Shangshufang, PKU canteen, THU canteen, HIT canteen}
More general definition of a choice set: X ✓ Rn

Young Joon Park (PHBS) Game Theory February/2023 8 / 31

Binary Relations

A binary relation R defined over a set S specifies some meaningful

relationship that may hold between a pair of elements of S.
eg) “ ”, “likes”, “ was born in the same province as”, etc.
Some ‘possible’ properties of binary relations are:
I Completeness: For any x, y 2 X , either xRy or y Rx.
I Transitivity: For any x, y , z 2 X , if xRy and y Rz, then xRz.
I Reflexivity: For all x 2 X , xRx.
I Irreflexivity: For any x 2 X , xRx does not hold.
I Symmetry: For any x, y 2 X , if xRy then y Rx.

Young Joon Park (PHBS) Game Theory February/2023 9 / 31

Making Choices using Preference Relations

DM making a choice of x0 2 X can imply x0 is ‘better’ than any other

x 2 X.
Formally, we write x0 % x for all x 2 X .
Here, the symbol % is a binary relation that indicates the DM
considers x0 is at least as good as x.
In short, we call % the ‘weak preference’ relation
I x0 % x: x0 is weakly preferred to x
With such definition, we can also define two other useful relations
such as
I Strict Preference: x x0 if and only if x % x0 and x0◆ %
◆x.
We say x is strictly preferred to x0 .
I Indi↵erence: x ⇠ x0 if and only if x % x0 and x0 % x.
We say x is indi↵erent with x0 .

Young Joon Park (PHBS) Game Theory February/2023 10 / 31

Some Properties of Preference Relations

Some ‘Possible’ Properties of Preference Relations

I Completeness
8 x, y 2 X , either x % y or y % x
I Transitivity
For any three elements x, y, z 2 X , if x % y and y % z, then x % z.
F Rationality of Preferences:
A preference relation % is called Rational if it satisfies Completeness
and Transitivity.
I Continuity (technical)
8 x 2 Rn+ the upper contour set and the lower contour set are closed in
Rn+ .

Young Joon Park (PHBS) Game Theory February/2023 11 / 31

Utility Representation

Utility Function
Definition
A real-valued function u : Rn+ ! R is called a utility function representing
the preference relation % if for all x, y 2 Rn+ , x % y () u(x) u(y).

Young Joon Park (PHBS) Game Theory February/2023 12 / 31

Existence of Utility Function

Theorem (Utility Representation Theorem)

If the preference ordering is complete, transitive, and continuous, then
there exists a continuous utility function f : X ! R which represents %.

With the utility representative theorem, we can model DM’s problem

in more comprehensive manner.

Young Joon Park (PHBS) Game Theory February/2023 13 / 31

DM’s New Problem

DM chooses to maximize a utility function u.

u specifies how much utility DM gets from each alternative:

u:X !R

Example: In R2+ , x1 = apples and x2 = bananas

I If, regardless of current (x1 , x2 ), 1 apple gives utility of 2 and 1 banana
gives utility of 5 to the DM, we can write the utility function as
u(x1 , x2 ) = 2x1 + 5x2 .
I If the DM always consumes 1 apple with 2 bananas that gives utility of
1. Also, he does not like any other combinations (any left over apples
or bananas are useless). Then we can write the utility function as
u(x1 , x2 ) = min[x1 , 12 x2 ].
I A more general form of a utility function would be u(x1 , x2 ) = Ax1↵ x2 .

Young Joon Park (PHBS) Game Theory February/2023 14 / 31

More Possible Properties of the Utility Function
Strictly Increasing Function
A function f : X ! R where X ⇢ Rn is strictly increasing if
(i) f (x) f (y) whenever x y, and
(ii) f (x) > f (y) whenever x y.
Quasiconcave Function
A function f : X ! R is quasiconcave if 8x, y 2 X and 8t 2 [0, 1],
f (tx + (1 t)y) min [f (x), f (y)]
(The convex combination is weakly greater than the lesser value.)
Strictly Quasiconcave Function
A function f : X ! R is strictly quasiconcave if 8x, y 2 X , such that
x 6= y and 8t 2 (0, 1),
f (tx + (1 t)y) > min [f (x), f (y)]
(The ‘strict’ convex combination is strictly greater than the lesser
value.)
Young Joon Park (PHBS) Game Theory February/2023 15 / 31
Choices under Uncertainty

All choices are made under some kind of uncertainty.

Sometimes it is useful to ignore uncertainty and focus on ultimate
choices.
Other times, we must model uncertainty explicitly.
Examples:
I Insurance markets.
I Financial markets.
I Game theory.

Young Joon Park (PHBS) Game Theory February/2023 16 / 31

Consequences

Two basic elements of expected utility theory: consequences (or

outcomes) and lotteries.
Finite set C of consequences.
Consequences are what the decision-maker ultimately cares about.
Example: “I get pneumonia, my health insurance company covers
most of the costs, but I have to pay a $500 deductible.”
Consumer does not choose consequences directly.

Young Joon Park (PHBS) Game Theory February/2023 17 / 31

Lotteries

Consumer chooses a lottery, p.

Lotteries are probability
P distributions over consequences:
p : C ! [0, 1] with c2C p(c) = 1.
Set of all lotteries is denoted by P.
Example: “A gold-level health insurance plan, which covers all kinds
of diseases, but has a $500 deductible.”
Makes sense because consumer assumed to rank health insurance
plans only insofar as lead to di↵erent probability distributions over
consequences.

Young Joon Park (PHBS) Game Theory February/2023 18 / 31

DM’s Choice under Uncertainty

What’s set of alternatives here, C or P?

Answer: P
Consumer does not choose consequences directly, but instead chooses
lotteries.
Just like the case of “Choice under Certainty”, assume the DM has a
rational preference relation % on P.

Young Joon Park (PHBS) Game Theory February/2023 19 / 31

Some Properties of Lotteries

Given two lotteries p and p 0 , the convex combination ↵p + (1 ↵)p 0

is the lottery defined by [↵p + (1 ↵)p 0 ](c) = ↵p(c) + (1 ↵)p 0 (c)
for all c 2 C .
One way to generate it:
I First, randomize between p and p 0 with weights ↵ and 1 ↵.
I Second, choose a consequence according to whichever lottery came up.
Such a probability distribution over lotteries is called a compound
lottery.
In expected utility theory, no distinction between simple and
compound lotteries: simple lottery ↵p + (1 ↵)p 0 and above
compound lottery give same distribution over consequences, so
identified with same element of P.

Young Joon Park (PHBS) Game Theory February/2023 20 / 31

Utility for Choice under Uncertainty

Whenever % is rational and continuous, it can be represented by

continuous utility function U : P ! R: p % q , U(p) U(q)
Intuitively, want more than this.
Want not only that consumer has utility function over lotteries, but
also that somehow related to “utility” over consequences.
Only care about lotteries insofar as a↵ect distribution over
consequences, so preferences over lotteries should have something to
do with “preferences” over consequences.

Young Joon Park (PHBS) Game Theory February/2023 21 / 31

Expected Utility

Best we could hope for is representation by utility function of

following form - Expected Utility Form:
Definition
A utility function U : P ! R has an expected utility form if there exists a
function u : C ! R such that
X
U(p) = p(c)u(c) for all p 2 P and c 2 C
c2C

In this case, the function U is called an a von Neumann-Morgenstern

utility function or expected utility function, and the function u is call a
Bernoulli utility function.

Young Joon Park (PHBS) Game Theory February/2023 22 / 31

More on Expected Utility

If preferences over lotteries happen to have an expected utility

representation, it’s as if consumer has a “utility function” over
consequences (and chooses among lotteries so as to maximize
expected “utility over consequences”).
X
U(p) = p(c)u(c)
c2C

U : P ! R is an example of a standard utility function.

Bernoulli utility function u : C ! R is not a standard utility function.
I Can’t have a “real” utility function on consequences, as consumer
never chooses among consequences.
I If preferences over lotteries happen to have an expected utility
representation, it’s as if consumer has a “utility function” over
consequences.

Young Joon Park (PHBS) Game Theory February/2023 23 / 31

Property of EU: Linearity in Probabilities

If U : P ! R is an expected utility function, then

U(↵P + (1 ↵)p 0 ) = ↵U(p) + (1 ↵)U(p 0 )

In fact, a utility function U : P ! R has an expected utility form i↵

this equation holds for all p, p 0 , and ↵ 2 [0, 1].

Young Joon Park (PHBS) Game Theory February/2023 24 / 31

Property of EU: Invariant to Affine Transformations
Suppose U : P ! R is an expected utility function representing
preferences %. Then, any increasing transformation of U also
represents %.
I The property of invariance to increasing transformation also applies to
utility function without uncertainty.
I Not all increasing transformations of U have expected utility form.
I Expected utility form is preserved for affine transformation.

Theorem
Suppose U : P ! R is an expected utility function representing
preferences %. Then V : P ! R is also an expected utility function
representing % i↵ there exist a, b > 0 such that

V (p) = a + bU(p) for all p 2 P.

P
If this is so, we also have V (p) = c2C p(c)v (c) for all p 2 P, where

v (c) = a + bu(c) for all c 2 C .

Young Joon Park (PHBS) Game Theory February/2023 25 / 31
Expected Utility

Preferences must be rational (complete and transitive) to have any

kind of utility representation.
Preferences on a compact and convex set must be continuous to have
a continuous utility representation.
Besides rationality and continuity, what’s needed to ensure that
preferences have an expected utility representation?
Definition
A preference relation % satisfies independence if, for every p, p 0 , p 00 2 P
and ↵ 2 (0, 1), p % p 0 , ↵p + (1 ↵)p 00 % ↵p 0 + (1 ↵)p 00 .

Can interpret as form of “dynamic consistency.”

Young Joon Park (PHBS) Game Theory February/2023 26 / 31

Expected Utility Theorem
Implication

Theorem (Expected Utility Theorem)

A preference relation % has an expected utility representation i↵ it satisfies
rationality, continuity, and independence.

It says given the assumptions stated, agents’ preference over lotteries

can be described in very simple way: by measuring the expected value
of utility.
This is a powerful result. -Once we know the utility number assigned
to all outcomes, the preference over uncertain lotteries can be
described by simply comparing the probability-weighted sum of the
utility number between di↵erent probability densities (lotteries).
It turns out that the expected utility framework is quite useful in the
analysis of attitude toward risk.

Young Joon Park (PHBS) Game Theory February/2023 27 / 31

Expected Utility Theorem
Indi↵erence Curves

If an agent is an expected utility maximizer, then she has straight and

parallel indi↵erence curves on the simplex.
The intuition on the straightness comes from the fact that

p ⇠ ↵p + (1 ↵)p ⇠ ↵p + (1 ↵)p 0

for all p ⇠ p 0 and all ↵ 2 [0, 1].

The indi↵erence curves should be parallel since for all p and p 0 such
that p ⇠ p 0 , and all p 00 ,

↵p + (1 ↵)p 00 ⇠ ↵p 0 + (1 ↵)p 00

for all ↵ 2 [0, 1].

Young Joon Park (PHBS) Game Theory February/2023 28 / 31

The Allais Paradox
Description of the Problem

Example (6.B.2)
There are three possible monetary prizes (so the number of outcomes is
N = 3):

First Prize Second Prize Third Prize

$ 25,000,000 $ 5,000,000 $0

The decision maker is subject to two choice tests. The first consists of a
choice between the lotteries p1 and p10 and the second consists of a choice
between the lotteries p2 and p20 :

p1 = (0, 1, 0) p10 = (.10, .89, .01)

p2 = (0, .11, .89) p20 = (.10, 0, .90).

Young Joon Park (PHBS) Game Theory February/2023 29 / 31

The Allais Paradox
The Typical Result and Its Implication

It is common for individuals to express the preferences p1 p10 and

p20 p2 .
The first choice means people preferred the sure payment of
$5,000,000 over the lottery that has 10% chance of receiving 5 times
more payment but also with small probability of getting nothing.
The second choice means that people preferred the lottery that pays
$25,000,000 with 10% probability to the lottery that has slightly
higher probability (11%) to receive significantly lower payment
($5,000,000).
By all means, the “typical” reaction is reasonable in many aspects:
I Even very small, a considerable probability of getting zero (very
disappointing) outcome could have prevented people to choose p10 over
p1 .
I A small probability (1%) advantage of desirable result could not beat
the 5 times increase in the result itself.

Young Joon Park (PHBS) Game Theory February/2023 30 / 31

The Allais Paradox
Independence Axiom

The result is called a “paradox”, however, since it does not satisfy

expected utility theorem (more precisely, it violates the independence
axiom).
p1 p10 implies

u(5M) > (.10)u(25M) + (.89)u(5M) + (.01)u(0).

Adding (.89)u(0) (.89)u(5M) to both sides, we get

(.11)u(5M) + (.89)u(0) > (.10)u(25M) + (.90)u(0).

Thus, an expected utility maximizer who prefers p1 to p10 should also

prefer p2 over p20 .
This observation could be also depicted on the simplex with three
outcomes.

Young Joon Park (PHBS) Game Theory February/2023 31 / 31