Sie sind auf Seite 1von 4

1.

Binary relations

1.1 Relations and orderings


Preference relations can seem counterintuitive at first, because you have prejudices
that relations should have certain properties. Eventually, you will assume that they
do in fact behave normally, but at first you will have to understand the weird ones.
So for now, throw out anything you think that preferences should do.
To talk about preference relations, first start with a set X of options or things. The
notation X X is used to indicate {(x, y): x X, y X} . For example, R R =
{(x, y): x R , y R} is the set of all vectors with both components real numbers, or
the real plane. This is often written R 2 for obvious notational reasons.
The next object is a binary relation, which is a subset of X X . This can be
written f X X . Its a bit odd at first (I think) to think of a relation as a subset,
but thats the convention. If (x, y) f , then we say x stands in relation f to y or we
might simply write x f y . In the context of preference maximization, the
interpretation is that x is at least weakly preferred to y.
Example:
X = {0,1,2,3} , and I prefer smaller numbers to larger numbers. In this case, the
relation is the set f = {(0,0),(0,1),(0,2),(0,3),(1,1),(1,2) , (1,3),(2,2),(2,3),(3,3)} .
When x f y and [y f x] , we write x f y and say x is strictly preferred to y.
Consider this a definition.
xf y

[ xf y & [ yf x]]

There is a similar notion of equivalence, called indifference for preference


relations. It will be defined as:
x~y

[x f y & y f x]

From these definitions, we can infer that x f y [ y ~ x] and that x ~ y [y f x] .


However, its not something that you should presume that you can assume.
The binary relation provides pairwise comparisons of some options: x is preferred to
y. For economics, we want for the preference relation to provide an ordering of the
available options, which means that we can talk about maximal (or minimal)
elements: x is the best option out of set X. Whats the difference?

Fall 2001 math for economic theory, page 2

Example: Consider a set of sports teams from different cities, X={Atlanta, Buffalo,
Chicago, Denver}. The binary relation is Team x beats or ties team y, and
f = {( A, B),(C, D)} . What is the top team out of X? We cant saysince team A never plays
team C or D, nor does team B play C or D, we cant make comparisons between them.
Example: The set of teams is the same and the binary relation is the same, but the set is
now f = {( A, B),( B,C),(C,D),(D,A)} . In other words, Atlanta beat Buffalo, Buffalo beat
Chicago, Chicago beat Denver, and then Denver beat Atlanta. Theres no team which is
superior to everyone else.
These are essentially the two types of complications which can arise when trying to
go from a binary relation to an ordering. The first is a failure of the relation to be
complete (meaning that not all pairs were compared), and the second is a failure of
transitivity.
Definition: The relation f on X is complete when for all x, y X , x f y or yf x .
Definition: The relation f on X is transitive when for all x, y,z X , if x f y and yf z ,
then x f z .
A preference relation that is both complete and transitive is called rational. On
page 7 of MWG, there is a good example of properties that you can infer from f
being rational. Again, its important to realize that these are exactly the sorts of
things that you should become accustomed to not presuming.
Prop.:
If f is rational, then f is irreflexive (that is, x f x never holds) and transitive.
Proof:
Irreflexiveness: Suppose that f is rational (that is, both complete and
transitive). Completeness means that for all x and y, either x f y or yf x . Let x = y.
Completeness implies that x f x. This rules out [x f x] , which is necessary for x f x,.
Transitivity: suppose not. Suppose that f is rational and that f is not transitive: that is, there
exist x, y,z X such that x f y , y f z but [x f z] . Since x f y xf y and y f z yf z ,
transitivity of f implies that x f z . Then for transitivity of f to fail,, it must be the case that
zf x . Combined with y f z , this implies that yf zf x y f x. But x f y [ y f x] . This
is a contradiction. Consequently, f must be transitive.
Lets step away from preference relations in economics, and return to binary
relations in math. The reason why a rational preference relation is so nice is that
Fall 2001 math for economic theory, page 3

its very closely related to something called a well-ordering. There are two parts to
defining this.
Definition: A binary relation f on a set X is a partial ordering (and the set X is
partially ordered if it satisfies:
1. f is reflexive: x X , x f x.
2. f is transitive: [x f y, y f z] x f z .
3. f is antisymmetric in the sense that [x f y, y f x] x = y .
The reason why this is called a partial ordering is that were not requiring that any
two elements can be either x f y or yf x . In that case, x and y would be said to be
noncomparablecompleteness fails.
A partial ordering without any
noncomparable pairs is called an ordering, and the set X would be called ordered
if f is an ordering on it. Sets which have maximal elements are known as wellordered. In terms of individual choice, this means that the person can use his
preference relation to find a preference maximization element in each set. For some
reason, this is usually stated in math as a first element or smallest element, but
the same goes for a last element or greatest element.
Definition: An ordered set X is well-ordered if every nonempty subset A X has a
smallest element; that is, some a such that a p a for every a A .
Example:
Example:

Every finite ordered set is well-ordered.


Every nonempty subset of a well-ordered set is itself well-ordered.

The first property has this interpretation: if preferences are rational, then every finite
set of options has a preference maximizing option. Unfortunately, this isnt true of
non-finite setsthe preference relation may not generate a well-ordering on these
sets. There need not be a failure. In fact, Zermelo gave us these happy results:
Theorem: Any set can be well-ordered.
Theorem: Given any set X , there is a choice function C such that C( B) is an element of
of B for every nonempty subset B X
The second result is often known as Zermelos axiom of choice. In this context, it says
that there always exist choice structures satisfying a basic requirement: that you
Fall 2001 math for economic theory, page 4

chose something out of the set offered to you. Later on, it shows up as a way to
prove the existence and uniqueness of subgame perfect strategies.

1.2

Exercises

Exercise 1.2.1:

Prove: If f is rational, then ~ is reflexive, transitive, and symmetric.

Exercise 1.2.2:

Prove: If f is rational, then x f y f z x f z .

Exercise 1.2.3:
Invent a binary relation satisfying the properties of a partial ordering
on the set of complex numbers.
Exercise 1.2.4:
Given subsets A, B of the space X, the relation is defined as A B
when all members of A are also in B. Verify that this is a partial ordering on 2 X . What is the
minimal element of 2 X , based on this ordering? What is the maximal element?
References:
MW&G, Chapter 1.
Kolomogorov & Fomin, Chapter 1.

Fall 2001 math for economic theory, page 5

Das könnte Ihnen auch gefallen