Economics N110, Game Theory in the Social Sciences

UC Berkeley, Summer 2012

Lecture 1
Instructor: Matt Leister

June 18, 2012

Table of contents
What is Game Theory? What is Game Theory? What to expect Mathematical Concepts Sets and Numbers Vectors and Vector Spaces Functions and Correspondences Optimization Sequences and Series Probability Implication

What is Game Theory?

the mathematical modeling of agent interaction: the study of strategy Examples: competition, cooperation, coordination, bargaining, even communication, and many more aspects of our every day lives can be modeled with game theory. a basic language to much of microeconomics and often applied in all subelds of economics the brain child of the late great John Von Neumann and Oskar Morgenstern Berkeley has a rich history in game theory: John Harsanyi (1920-2000, Cal professor, and Nobel Laureate) developed the theory of games with imperfect information.

What to expect from this course

1. a sound understanding of essential concepts 2. a roadmap of the theorys models and concepts 3. the tools to continue your studies in game theory and microeconomics 4. a survey of how game theory has been applied in dierent social science elds (e.g. sociology, political science, public policy, psychology, and economics)

Mathematical Concepts: Sets

A set is a collection of things or elements(e.g. types of fruit, people, choices, variables or numbers). Sets can be nite (e.g. days of the week) or innite (e.g. time(s) it takes to drive to San Francisco). nite sets often denoted as {, , . . . , } innite but orderable sets usually denoted as {, , . . .} this course: 1) sets of choices (actions, strategies) 2) sets of states of the world (e,g, {rainy, sunny}) Write x X to denote that the variable x is an element in the set X .

Notation
x X ... (short: x ...), means for all x in X.... x X ... (short: x ...), means there exists an x in X.... Take X and Y subsets in Z . This is written X Z and Y Z. X Y means the set of all elements either in X or in Y X Y means the set of all elements in both X and Y X c means the set of all elements not in x (but in Z) X \Y means the set of all elements in X but not in Y (equal to X Y c ) Fact(s): 1) (X Y )c = X c Y c , 2) (X Y )c = X c Y c (De Morgans Law)

Sets of Sets and Sets of Numbers

A set can be composed of sets (each element is a set itself), and we usually denote these types of sets in script (e.g. X ) The set of natural numbers (e.g. 1, 2, . . .) are denoted N The set of integers (e.g. 5, 0, 3) are denoted Z The set of real numbers (e.g. 5.23, , 2) are denoted R

Vectors and Vector Spaces

Taking a set X composed of n R sets X1 ,...,Xn , a vector can be dened as a bundle of n elements, one taken from each of the sets in X . This course: well refer to vectors as proles. The set of all such vectors is denoted X1 X2 ... Xn (or n i =1 Xi ). If Xi = R, then we denote
n i =1 Xi

as Rn .

Rn is a vector space because we can dene component wise addition (and also inner products, which are not needed in this course). Example: x1 = (3.2, 6), x2 = (1.3, 2) (both in R2 ), with x1 + x2 = (1.9, 8)

Functions and Correspondences

Functions are mappings from one set to unique elements in another set (possibly equal to the rst set). Precisely, a function f () maps from a set X (the domain) to a set Y (the range); this is denoted f : X Y . For each x X , f gives one, and only one, y Y (denoted f (x )). This course: the functions well be dealing with are objective functions (next class) When a mapping gives more than one value for some elements, its referred to as a correspondence.

Example
A function (left) and a correspondence (right), each from R to R, can be graphed:

Continuity
A function from Rn to R is continuous if for each x X , f (x ) is exceedingly close to f (x ) as x approaches x . Examples:

Optimization

Important question: when a function f : X R, then which x X (possibly multiple x ) give the largest f (x )? The set of all such x is written: argmaxx X f (x ) The value of f at any such x is written: maxx X f (x )

Optimization
Fact: when X is nite, there will always be at least one x X that is in arg maxx X f (x ) However, when X is innite, it may be that arg maxx X f (x ) is empty ( denotes the empty set). Example: X =]0, 1[ (]a, b [ is set of integers between a and b , not including a and b ), with f (x ) = x . Fact: when X is a closed and bounded subset of Rn (i.e. X includes its boundry and all elements of X have nite components), and f () is continuous, then arg maxx X f (x ) is non-empty (due to Karl Weierstrass).

Example 1
This function from R2 to R has a maximum:

Figure : multi-modal density function

Example 2
Parabolas of the form y = f (x ) = ax 2 + bx + c (with a > 0) b are optimized at x = 2a :

a > 0, x = a < 0, x =

b 2a b 2a

gives a minimum; gives a maximum.

Example 2

Fact: for any f (), arg max f (x ) = arg min f (x )

x x

Fact: for any f (), max f (x ) = min f (x )

x x

Sequences
A sequence is an ordered innite set, usually denoted {x1 , x2 , ...}, or {xi } in short. A geometric sequence is of the form {1, r , r 2 , r 3 , ...}, r R (with r 0 = 1). a series is a sequence generated by the cumulative sums of a sequence. That is, the series of {x1 , x2 , ...} is the sequence {x1 , x1 + x2 , x1 + x2 + x3 , ...}. Fact: the Tth element of the series of a geometric sequence {1, r , r 2 , r 3 , ...} is: 1 + r + r 2 + ... + r T = 1 r T +1 1r
1 1r

Fact: if r (0, 1), then 1 + r + r 2 + ... =

Probability
We can dene the set of relevant states of the world and denote this set , with some probabilility distribution capturing the llikelihood of each event. If is nite, this means that we can nd Pr( ) [0, 1], , with Pr( ) = 1. We have Pr() = 0 and Pr() = 1. This course: nite random variables over action sets. A random variable (in this class) will be a function x from to R. With Pr(), we write Pr( x = x ) |x Pr( ). ( )=x A cumulative distribution F of x maps from R to [0, 1], and is dened as F (x ) |x x = x ). ( )x Pr( ) = x x Pr(

Conditional Probability

An event is some subset of . The probability of an event E is the sum of probabilities of the states within E : Pr(E ) E Pr( ). The conditional probability of the event E on the event F is the sum of probabilities of states in both E and F divided by and F ) the probability of F : Pr(E |F ) Pr(E . Pr(F ) This course: conditional probabilities will be used to model posterior beliefs (games with imperfect information)

Conditional Probability
Example: gene testing

Figure : relative densities

Bayes Rule

Using the above denitions, we can write: Pr(E and F ) = Pr(E |F ) Pr(F ) = Pr(F |E ) Pr(E ) Rearranging the second equality gives: Pr(E ) Pr(F |E ) Pr(E ) Pr(F |E ) = Pr(F ) Pr(F ) Pr() Pr(E ) Pr(F |E ) = Pr(E ) Pr(F |E ) + Pr(E c ) Pr(F |E c )

Pr(E |F )

Bayes Rule

A partition of is a collection of disjoint subsets of with union equal to . Taking nite partition {Ej }n j =1 of , the above expression can be more generally written: Pr(Ek |F ) = Pr(Ek ) Pr(F |Ek ) n j =1 Pr(F |Ej ) Pr(Ej )

Implication

When a fact, event, or claim A implies another fact, event, or claim B , we write A B (or B A). When A B and B A (read A if and only i B ), we write A B or A i B . Alternative terminology: A B is the same as A is sucient for B and B is necessary for A