Structural Properties of Utility Functions

Walrasian Demand

Econ 2100 Fall 2017

Lecture 4, September 7

Outline
1 Structural Properties of Utility Functions
1 Local Non Satiation
2 Convexity
3 Quasi-linearity
2 Walrasian Demand
 From Last Class
Denition
The utility function u : X ! R represents the binary relation % on X if
x % y , u(x) u(y ):

Theorem (Debreu)
Suppose X Rn . A binary relation % on X is complete, transitive, and continuous
if and only if it admits a continuous utility representation u : X ! R.

Today we study connections between a utility function and the underlying

preference relation it represents.

Structural Properties of Utility Functions

The main idea is to understand the connection between properties of
preferences and characteristics of the utility function that represents them.

NOTATION:
We assume X = Rn .
If xi yi for each i, we write x y.
 Local Non Satiation
Denition
A preference relation % is locally nonsatiated if for all x 2 X and " > 0, there exists
some y such that ky xk < " and y x.

For any consumption bundle, there is always a nearby bundle that is strictly
preferred to it. (Draw an example)

Denition
A utility function u : X ! R is locally nonsatiated if it represents a locally
nonsatiated preference relation %; that is, if for every x 2 X and " > 0, there exists
some y such that ky xk < " and u(y ) > u(x).

Example: The lexicographic preference on R2 is locally nonsatiated

Fix (x1 ; x2 ) and " > 0.
Then (x1 + "2 ; x2 ) satises k(x1 + "2 ; x2 ) (x1 :x2 )k < "
"
and (x1 + 2 ; x2 ) (x1 ; x2 ).
 Local Non Satiation and Strict Monotonicity

Proposition
If % is strictly monotone, then it is locally nonsatiated.

Proof.
Let x be given, and let y = x + n" e, where e = (1; :::; 1).
Then we have yi > xi for each i.
Strict monotonicity implies that y x.
Note that v
u n
uX " 2 "
jjy xjj = t = p < ":
n n
i =1

Thus % is locally nonsatiated.

 Shapes of Functions

Denitions
Suppose C is a convex subset of X . A function f : C ! R is:
concave if
f ( x + (1 )y ) f (x) + (1 )f (y )
for all 2 [0; 1] and x; y 2 C ;
strictly concave if
f ( x + (1 )y ) > f (x) + (1 )f (y )
for all 2 (0; 1) and x; y 2 X such that x 6= y ;
quasiconcave if
f (x) f (y ) ) f ( x + (1 )y ) f (y )
for all 2 [0; 1];
strictly quasiconcave if
f (x) f (y ) and x 6= y ) f ( x + (1 )y ) > f (y )
for all 2 (0; 1).
 Convex Preferences
Denitions
A preference relation % is
convex if
x %y ) x + (1 )y % y for all 2 (0; 1)
strictly convex if
x % y and x 6= y ) x + (1 )y y for all 2 (0; 1)

Convexity says that taking convex combinations cannot make the decision
maker worse o.
Strict convexity says that taking convex combinations makes the decision
maker better o.

Question
What does convexity imply for the utility function representing %?
 Convex Preferences: An Example

Let % on R2 be dened as x % y if and only if x1 + x2 y1 + y2 is convex

Proof: Suppose x % y , i.e. x1 + x2 y1 + y2 , and x 2 (0; 1).
Then
x + (1 )y = ( x1 + (1 )y1 ; x2 + (1 )y2 )
So,
[ x1 + (1 )y1 ] + [ x2 + (1 )y2 ] = [x1 + x2 ] + (1 )[y1 + y2 ]
| {z }
y1 +y2
[y1 + y2 ] + (1 )[y1 + y2 ]
= y1 + y2 ;
proving x + (1 )y % y .
This is not strictly convex, because (1; 0) % (0; 1) and (1; 0) 6= (0; 1) but
1 1 1 1
(1; 0) + (0; 1) = ( ; ) % (0; 1):
2 2 2 2
 Convexity and Quasiconcave Utility Functions

Convexity is equivalent to quasi concavity of the corresponding utility function.

Proposition
If u represents %, then:
1 % is convex if and only if u is quasiconcave;
2 % is strictly convex if and only if u is strictly quasiconcave.

Convexity of % implies that any utility representation is quasiconcave, but not

necessarily concave.

Proof.
Question 5b. Problem Set 2, due next Tuesday.
 Quasiconcave Utility and Convex Upper Contours
Proposition
Let % be a preference relation on X represened by u : X ! R. Then, the upper
contour set is a convex subset of X if and only if u is quasiconcave.

Proof.
Suppose that u is quasiconcave.
Fix z 2 X , and take any x ; y 2% (z).
Wlog, assume u(x ) u(y ), so that u(x ) u(y ) u(z), and let 2 [0; 1].
By quasiconcavity of u,
u(z) u(y ) u( x + (1 )y );
so x + (1 )y % z.
Hence x + (1 )y belongs to % (z), proving it is convex.
Now suppose the better-than set is convex.
Let x ; y 2 X and 2 [0; 1], and suppose u(x ) u(y ).
Then x % y and y % y , and so x and y are both in % (y ).
Since % (y ) is convex (by assumption), then x + (1 )y % y .
Since u represents %,
u( x + (1 )y ) u(y )
Thus u is quasiconcave.
 Convexity and Induced Choices
Proposition
If % is convex, then C% (A) is convex for all convex A.
If % is strictly convex, then C% (A) has at most one element for any convex A.

Proof.
Let A be convex and x; y 2 C% (A).
By denition of C% (A), x % y .
Since A is convex: x + (1 )y 2 A for any 2 [0; 1].
Convexity of % implies x + (1 )y % y .
By denition of C% , y % z for all z 2 A.
Using transitivity, x + (1 )y % y % z for all z 2 A.
Hence, x + (1 )y 2 C% (A) by denition of induced choice rule.
Therefore, C% (A) is convex for any convex A.
Now suppose there exists a convex A for which C% (A) 2.
Then there exist x ; y 2 C% (A) with x 6= y .
Since A is convex, x + (1 )y 2 A for all 2 (0; 1).
Since x % y and x 6= y , strict convexity implies x + (1 )y y , but this
contradicts the fact that y 2 C% (A).
 Quasi-linear Utility
Denition
The function u : Rn ! R is quasi-linear if there exists a function v : Rn 1
!R
such that u(x; m) = v (x) + m.

We usually think of the n-th good as money (the numeraire).

Proposition
The preference relation % on Rn admits a quasi-linear representation if and only
1 (x; m) % (x; m0 ) if and only if m m0 , for all x 2 Rn 1
and all m; m0 2 R;
2 (x; m) % (x ; m ) if and only if (x; m + m ) % (x ; m + m00 ), for all x 2 Rn
0 0 00 0 0 1

and m; m0 ; m00 2 R;
3 for all x; x 0 2 Rn 1
, there exist m; m0 2 R such that (x; m) (x 0 ; m0 ).

1 Given two bundles with identical goods, the consumer always prefers the one
with more money.
2 Adding (or subtracting) the same monetary amount does not change rankings.
3 Monetary transfers can always be used to achieve indierence.
Proof.
Question 5c. Problem Set 2, due next Tuesday.
 Quasi-linear Preferences and Utility

Proposition
Suppose that the preference relation % on Rn admits two quasi-linear
representations: v (x) + m, and v 0 (x) + m, where v ; v 0 : Rn 1 ! R. Then there
exists c 2 R such that v 0 (x) = v (x) c for all x 2 Rn 1 .

Proof.
Exercise
 Homothetic Preferences and Utility

Homothetic preferences are also useful in many applications, in particular for

aggregation problems and macroeconomics.

Denition
The preference relation % on X is homothetic if for all x; y 2 X ,
x y) x y for each >0

Proposition
The continuous preference relation % on Rn is homothetic if and only if it is
represented by a utility function that is homogeneous of degree 1.

r
A function is homogeneous of degree r if f ( x) = f (x) for any x and > 0.

Proof.
Question 5d. Problem Set 2, due next Tuesday.
 Demand Theory

Main Questions
Suppose the consumer uses her income to purchase goods (commodities) at
the exogenously given prices:
What are the optimal consumption choices?
How do they depend on prices and income?

Typically, we answer this questions solving a constrained optimization problem

using calculus.
That means the utility function must be not only continuous, but also
dierentiable.
Dierentiability is not a property we can derive from preferences.
That is, however, not necessary and we can talk about optimal choices even
when preferences are not necessarily represented by a utility function.
 Budget Set
First, we dene what a consumer can buy.

Denition
The Budget Set B(p; w ) Rn at prices p and income w is the set of all aordable
consumption bundles and is dened by
B(p; w ) = fx 2 Rn+ : p x w g:

This is the set of consumption bundles the consumer can choose from. She
cannot purchase consumption bundles outside of this set.
Implicit assumptions: goods are perfectly divisible; consumption is non
negative; the total price of consumption cannot exceed income; prices are
linear. Think of possible violations.

Exercise
Suppose w = $100. There are two commodities, electricity and food. Each unit of
food costs $1. The rst 20Kwh electricity cost $1 per Kwh, but the price of each
incremetal unit of electricity is $1:50 per Kwh. Write the consumers budget set
formally and draw it.
 Walrasian Demand
Main Idea
The optimal consumption bundles are those that are preferred to all other
aordable bundles.

Denition
Given a preference relation %, the Walrasian demand correspondence
x : Rn++ R+ ! Rn+ is dened by
x (p; w ) = fx 2 B(p; w ) : x % y for any y 2 B(p; w )g:

By denition, for any x 2 x (p; w )

x %x for any x 2 B(p; w ):
Walrasian demand equals the induced choice rule given the preference relation
% and the available set B(p; w ):
x (p; w ) = C% (B(p; w )):
More implicit assumptions: income is non negative; prices are strictly positive.
 Walrasian Demand With Utility

Although we do not need the utility function to exist to dene Walrasian

demand, if a utility function exists there is an equivalent denition.

Denition
Given a utility function u : Rn+ ! R, the Walrasian demand correspondence
x : Rn++ R+ ! Rn+ is dened by
x (p; w ) = arg max u(x) where B(p; w ) = fx 2 Rn+ : p x w g:
x 2B (p;w )

As before,
x (p; w ) = C% (B(p; w )):
and for any x 2 x (p; w )
u(x ) u(x) for any x 2 B(p; w ):
We can derive some properties of Walrasian demand directly from assumptions
on preferences and/or utility.
Walrasian Demand Is Homogeneous of Degree Zero

Proposition
Walrasian demand is homogeneous of degree zero: for any >0
x ( p; w ) = x (p; w )

Proof.
For any > 0,
B( p; w ) = fx 2 Rn+ : p x w g = fx 2 Rn+ : p x w g = B(p; w )
because is a scalar
Since the constraints are the same, the optimal choices must also be the
same.
 The Consumer Spends All Her Income
This is sometimes known as WalrasLaw for individuals

Proposition (Full Expenditure)

If % is locally nonsatiated , then
p x =w for any x 2 x (p; w )

Proof.
Suppose not.
Then there exists an x 2 x (p; w ) with p x < w
Find some y such that
ky xk < " with " > 0 and p y w
(why does such a y always exist?)
By local non satiation, this implies y x contradicing x 2 x (p; w ).
 Walrasian Demand Is Convex
Proposition
If u is quasiconcave, then x (p; w ) is convex.
If u is strictly quasiconcave, then x (p; w ) is unique.

Same as before (u (stricly) quasiconcave means % (strictly) convex).

Proof.
Suppose x; y 2 x (p; w ) and pick 2 [0; 1].
First convexity: need to show x + (1 )y 2 x (p; w ).
x % y by denition of x (p; w ).
u is quasiconcave, thus % is convex and x + (1 )y % y .
y % z for any z 2 B (p; w ) by denition of x (p; w ).
Transitivity implies x + (1 )y % z for any z 2 B (p; w ); thus
x + (1 )y 2 x (p; w ).
Now uniqueness.
x ; y 2 x (p; w ) and x 6= y imply x + (1 )y y for any 2 (0; 1) because
u is strictly quasiconcave (% is strictly convex).
Since B (p; w ) is convex, x + (1 )y 2 B (p; w ), contradicting
y 2 x (p; w ).
 Walrasian Demand Is Non-Empty and Compact

Proposition
If u is continuous, then x (p; w ) is nonempty and compact.

We already proved this as well.

Proof.
Dene A by
A = B(p; w ) = fx 2 Rn+ : p x wg
This is a closed and bounded (i.e. compact, set) and
x (p; w ) = C% (A) = C% (B(p; w ))
where % are the preferences represented by u.
Then x (p; w ) is the set of maximizers of a continuous function over a
compact set.
 Walrasian Demand: Examples
How do we nd the Walrasian Demand?
Need to solve a constrained maximization problem, usually using calculus.

Question 6, Problem Set 2; due next Tuesday.

For each of the following utility functions, nd the Walrasian demand
correspondence. (Hint: pictures may help)
Q
n
1 u(x) = xi i with i > 0 (Cobb-Douglas).
i =1
2 u(x) = minf 1 x1 ; 2 x2 ; :::; n xn g with i > 0 (generalized Leontief).
Pn
3 u(x) = i =1 i xi for i > 0 (generalized linear).
Pn 1
4 u(x) = i =1 i xi (generalized CES).

Can we do the second one using calculus?

How about the third? Do we need calculus?
Constant elasticity of substitution (CES) preferences are the most commonly
used homothetic preferences. Many preferences are a special case of CES.
 An Optimization Recipe

How to solve max f (x) subject to gi (x) 0 with i = 1; ::; m

1 Write the Langrange function L : Rn Rm ! R as
Xm
L (x; ) = f (x) i gi (x)
i =1
2 Write the First Order Conditions:
n 1
z }| { m
X
rL (x; ) = rf (x) i rgi (x) = 0
i =1
| {z }
@f (x) P
m
@g i (x)
@xj i @x =0 for all j =1;::;n
i =1 j

3 Write constraints, inequalities for , and complementary slackness conditions:

gi (x) 0 with i = 1; ::; m
i 0 with i = 1; ::; m
i gi (x) = 0 with i = 1; ::; m
4 Find the x and that satisfy all these and you are done...hopefully.
 The Recipe In Action: Cobb-Dougals Utility
Compute Walrasian demand when the utility function is u(x1 ; x2 ) = x1 x21
Here x (p; w ) is the solution to
max x1 x21
x1 ;x2 2fp 1 x1 +p 2 x2 w ; x1 0, x2 0g

1 The Langrangian is
L (x; ) = x1 x21 w (p1 x1 + p2 x2 w) ( 1 x1 ) ( 2 x2 )
2 The First Order Conditions for x is:
0 1 0 u(x1 ;x2 )
1
x1 1 x21 w p1 + 1 x1 w p1 + 1
rL (x; ) = @ A =B
@
C
A=0
| {z }
2 1
(1 ) x 1 x 2 w p 2 + 2 (1 ) u(xx12;x2 ) w p2 + 2

3 The constraints, inequalities for , and complementary slackness are:

p 1 x1 + p 2 x2 w 0 x1 0, and x2 0
w 0, 1 0, and 2 0
w (p x
1 1 + p x
2 2 w ) = 0, x
1 1 = 0, and x
2 2 = 0
4 Find a solution to the above (easy for me to say).
 The Recipe In Action: Cobb-Dougals Utility
Compute Walrasian demand when the utility function is u(x1 ; x2 ) = x1 x21
We must solve:
u(x1 ;x2 )
x1 w p1+ 1 = 0 and (1 ) u(xx12;x2 ) w p2 + 2 =0
p1 x1 + p2 x2 w 0
and w 0; 1 0; 2 0
x1 0; x2 0
w (p1 x1 + p2 x2 w ) = 0 and 1 x1 = 0; 2 x2 =0

x (p; w ) must be strictly positive (why?), hence 1 = 2 = 0.

The budget constraint must bind (why?), hence w 0.
Therefore the top two equalities become
u(x1 ; x2 ) = w p1 x1 and (1 ) u(x1 ; x2 ) = w p 2 x2
Summing both sides and using Full Expenditure we get
u(x1 ; x2 ) = w (p1 x1 + p 2 x2 ) = ww
Substituting back then yields
1
w (1 )w 1
x1 (p; w ) = ; x2 (p; w ) = ; and w =
p1 p2 p1 p2
 Next Week

More Properties of Walrasian Demand.

Indirect Utility.
Comparative Statics.
Expenditure Minimization.