Vous êtes sur la page 1sur 67

Microeconomics 1

November 16, 2015


1

Disclaimer: These notes is my own condensed version of Chapters 1,2,3 and 5 in Jehle
and Reny. It is oered to you for your convenience but should not be seen as a replacement
of the material in the handbook.
Contents

1 Consumer behaviour 4
1.1 Primitive notions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2 Preferences and utility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2.1 Preference relation . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2.2 Utility function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.3 The consumers problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.4 Indirect utility and expenditure . . . . . . . . . . . . . . . . . . . . . . . . 12
1.4.1 The indirect utility function . . . . . . . . . . . . . . . . . . . . . . 12
1.4.2 The Expenditure function . . . . . . . . . . . . . . . . . . . . . . . 14
1.4.3 Relations between the two . . . . . . . . . . . . . . . . . . . . . . . 16
1.5 Properties of Consumer Demand . . . . . . . . . . . . . . . . . . . . . . . 18
1.5.1 Relative prices and income . . . . . . . . . . . . . . . . . . . . . . . 18
1.5.2 Income and substitution eects . . . . . . . . . . . . . . . . . . . . 19
1.5.3 Some elasticity relations . . . . . . . . . . . . . . . . . . . . . . . . 21
1.6 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

2 Topics in Consumer Theory 23


2.1 Duality, a closer look . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.1.1 Expenditure and consumer preferences . . . . . . . . . . . . . . . . 23
2.1.2 Convexity and monotonicity . . . . . . . . . . . . . . . . . . . . . . 25
2.1.3 Indirect utility and consumer demand . . . . . . . . . . . . . . . . . 25
2.2 Integrability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.3 Revealed preferences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.4 Uncertainty . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.4.1 Preferences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.4.2 Von Neumann-Morgenster utility . . . . . . . . . . . . . . . . . . . 32
2.4.3 Risk aversion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

3 Theory of the rm 37
3.1 Primitive notions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.2 Production . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.2.1 Returns to scale and varying proportions . . . . . . . . . . . . . . . 39
3.3 Cost . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

2
CONTENTS 3

3.4 Duality in production . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43


3.5 The competitive rm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.5.1 Prot maximisation . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.5.2 The prot function . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.6 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

5 General Equilibrium 49
5.1 Equilibrium in Exchange . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
5.2 Equilibrium in competitive market systems . . . . . . . . . . . . . . . . . . 50
5.2.1 Existence of equilibrium . . . . . . . . . . . . . . . . . . . . . . . . 50
5.2.2 Eciency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
5.3 Equilibrium in production . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
5.3.1 Producers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
5.3.2 Consumers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
5.3.3 Equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
5.3.4 Welfare . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
5.4 Contingent plans . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
5.4.1 Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
5.4.2 Uncertainty . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
5.4.3 Walrasian equilibrium with contingent commodities . . . . . . . . . 62
5.5 Core and equilibria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
5.5.1 Replica economies . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
Chapter 1

Consumer behaviour

1.1 Primitive notions


There are four building blocks in any model of consumer choice. They are the

the consumption set,

the feasible set,

the preference relation,

the behavioural assumption.

By specifying the form each of these takes in a given problem, many dierent situations
involving choice can be formally described and analysed.

We let the consumption set, X, represent the set of all alternatives, or complete
consumption plans, that the consumer can conceive whether some of them will be
achievable in practice or not. In a consumption setting it is standard
[ to take X to be
]
the set R+ , where is the number of goods. An element x X, x = x1 x2 . . . x

is then called a consumption bundle. The jth element of x, xj species the amount
of good j in the bundle x.
The minimal requirements on the consumption set are

X Rn+
X is closed.
X is convex.
0 X.

The choice set is the set of all feasible bundles. We denote this by B. It species
what the DM can actually choose from. Of course, B must be a subset of X.

4
CHAPTER 1. CONSUMER BEHAVIOUR 5

The preferences, R of the DM species what the decision maker cares about.

Behavioural assumption expresses the guiding principle the consumer uses to


make nal choices and so identies the ultimate objectives in choice. It is supposed
that the consumer seeks to identify and select an available alternative that is most
preferred in the light of his personal tastes.
A DM is said to behave rational if from every choice set B she never chooses an
alternative x when there is another alternative, say z, which is also in the choice set
B which is better according to her preferences. However, before we can specify what
a DM will choose, it is important to describe how her preferences look like.

1.2 Preferences and utility


1.2.1 Preference relation
Consumer preferences are characterised axiomatically. The axioms of consumer choice
are intended to give formal mathematical expression to fundamental aspects of consumer
behaviour and attitudes towards the objects of choice.

Preferences are dened as a binary relation R over the set of alternatives X (the hand-
book uses the notation for R).

R X X.

If (x, y) R, we say that the DM considers x at least as good as y. We write this also as
xRy.
We can impose several properties on this preference relations. The two most basic ones
are completeness and transitivity.

Axiom 1 The preference relation is complete if for all x, y X either xRy or yRx.
Completeness requires that the DM is able to rank any two alternatives in the choice space.

Axiom 2 A preference relation is transitive if for any three alternatives x, y and z X,


xRy and yRz implies xRz; if x is at least as good as y and y is at least as good as z then
x is at least as good as z.

Denition 1.1

A preference relation is a transitive and complete relation over X.


CHAPTER 1. CONSUMER BEHAVIOUR 6

Denition 1.2

The strict preference relation P ,

xP y if xRy and not yRx.

Denition 1.3

The indierence relation, I, is dened as,

xIy if xRy and yRx.

Denition 1.4

Let us dene the following sets,

R(x) = {y X|yRx}, P (x) = {y X|yP x},


I(x) = {y X|yIx}, (x)I = {y X|xIy},
(x)R = {y X|xRy}, (y)P = {y X|xP y}.

The set R(x) contains all alternatives that are at least as good as x. It is also called the
upper contour set of x. The set (x)R is also called the lower contour set. It collects
all bundles that are not better than x. The set I(x) is called the indierence curve
through x. Given that I is symmetric, we have I(x) = (x)I. Further, it is easy to see that,

R(x) = P (x) I(x),


I(x) = R(x) (x)R.
P (x) = X (x)R,
P (x) (x)P = .

Axiom 3 The preference relation R is said to be continuous if for all x X, the sets
R(x) and (x)R are closed (i.e. they contain their boundaries).
Continuity embeds the idea that people do not have sudden preference reversals. If con-
tinuity is satised, then I(x) = R(x) (x)R is also closed, as it is the intersection of two
closed sets and P (x) = X (x)R is open as it is the complement of a closed set.

Axiom 4 A preference relation R satisfies locally non-satiation if for all x X and


all > 0, there is an alternative y B (x) X such that yP x where B (x) is the open
CHAPTER 1. CONSUMER BEHAVIOUR 7

ball of radius around x. In other words, one can always find an alternative arbitrarily
close to x which is strictly preferred to x.

Axiom 4 A preference relation is strictly monotone if y x implies yRx and y x


implies yP x (here y x if yi > xi for all i ). Strict monotonicity states that the
decision maker prefers having strictly more of every good.
If a preference relation is strictly monotone, then it is locally non-satiated.
In addition, Monotonicity implies downward sloping indierence curves: if y x, then
y / I(x).

Axiom 5 A preference relation R is convex if the upper contour sets R(x) are convex
sets. In other words, if yRx, then for all t [0, 1], (ty + (1 t)x)Rx.

Axiom 5 A preference relation R is strictly convex if x = y and yRx implies that for
all t ]0, 1[, (ty + (1 t)x)P x.
Convexity is equivalent to the law of decreasing marginal rates of substitution. The
more you have of a certain good, the less you are willing to give up of another good to
have one more additional unit of that good (and keep utility the same).

1.2.2 Utility function


Denition 1.5

A utility function representing R is a real valued function u : X R such that


for all x, y X,
xRy u(x) u(y).

Observe that this also implies that xP y u(x) > u(y). Indeed if xP y and u(y)
u(x), then yRx which is a contradiction.
Two questions are important. When does there exist a utility function? and, if it exists,
in what sense is it unique? The next result gives the conditions for existence.

Theorem 1.1

If R is complete, transitive, strictly monotone, and continuous then there exists


a continuous utility function that represents R.
CHAPTER 1. CONSUMER BEHAVIOUR 8
[ ]
Proof. step I Let e = 1 1 1 . . . 1 be the vector of ones and for all x, dene the
number u(x) by,
u(x)eIx.

step II We need to show that u(x) is a continuous utility function. To start, let us rst
show that u(x) is well dened (i.e. u(x) exists and is uniquely dened).
Fix x and let A = {t R|teRx} and B = {t R|xRte}. By continuity of R,
both sets A and B are closed. By strict monotonicity, the sets A takes on the form
[r, [ and the set B = [0, k]. By completeness, we have that A B = R+ . This
establishes that r = k, so u(x) exists for all x. In order to show that it is unique, let,
xIae, xIbe and a > b. However, by strict monotonicity, we have that aeP be which
implies x I ae P be I x, a contradiction.

step III We still need to show that u(x) is a utility function. Let xRy. Then u(x)e I x R y I u(y)e,
so by transitivity u(x)e R u(y)e. By monotonicity, this gives u(x) u(y). If x P y,
we have u(x)e P u(y)e which means, by strict monotonicity, that u(x) > u(y).

step IV Finally, we need to show that u is continuous. in order to show this, it suces
to show that for every two numbers a, b R, with a < b, u 1(]a, b[) is an open set.
Indeed,

u1 (]a, b[) = {x X|a < u(x) < b},


= {x X|be P xP ae},
= {x X|x (be)P P (ae)}.

The intersection of two open sets is open which shows the result so u1 (]a, b[) is also
open.

Remark It can be shown that every complete, transitive and continuous preference relation
can be represented by a continuous utility function, so we do not really need monotonicity.
However the proof of this would be considerably more complicated.

Theorem 1.2

The functions u(x) and v(x) are two (continuous) utility functions for the same
preference relation R if and only if there is a strictly increasing (continuous)
function g : R R such that u(x) = g(v(x)).
CHAPTER 1. CONSUMER BEHAVIOUR 9

Theorem 1.3

R is strictly monotone i u(x) is strictly increasing.

R is convex i u is quasi-concave.

R is strictly convex i u is strictly quasi-concave. for x, y X and


t [0, 1],

u(tx + (1 t)y) min{u(x), u(y)}.

with a strict inequality if x = y and t ]0, 1[.

The set I(x) is equivalent to the set {y|u(x) = u(y)}. If there are only two goods, then,
by the implicit function theorem, we can dene the function x2 (x1 ) by u(x1 , x2 (x1 )) = cte .
The function x2 (x1 ) traces out the indierence curve in x1 , x2 space. If u is dierentiable,
then,

u(x1 , x2 (x1 )) = a,
u(x1 , x2 ) u(x1 , x2 ) dx2 (x1 )
+ = 0,
x1 x2 dx1
u(x1 ,x2 )
dx2
= u(x
x1
1 ,x2 )
= M RS1,2 .
dx1
u2

The left hand side is called the marginal rate of substitution. It measures the slope of
the indierence curve through the point x. It is dened as minus the ratio of the marginal
utilities of the two goods. It measures how much more of good 2 you need to give up in
order to have an addition unit of good one and be equally well o.
It can be generalized to more than two goods by setting,
u(x)
xi
M RSi,j = u(x)
.
xj

Observe that the marginal rate of substitution does not depend on the cardinalisation
chosen for the utility function.

1.3 The consumers problem


Assumption 1.2 Preferences are continuous, complete, transitive, strictly monotonic
and strictly convex.
CHAPTER 1. CONSUMER BEHAVIOUR 10

The budget set denes the bundles that are feasible to the decision maker. In a con-
sumption setting, these are all the goods
[ that] the consumer can possibly buy. Consider a
row vector of prices p R++ , p = p1 , . . . , p and an income level m (the book uses the
n

notation y for budget). The budget set is then dened as the set,

B = {x X|p x m}.

This set is closed and bounded, so it is compact.


If the decision maker is rational, then he will choose the best bundle from all feasible. In
other words, she will buy the bundle that maximizes her utility function given the budget
set.

max u(x) s.t. p x m.


xX

If the utility function is continuous, then the utility maximization problem has a solution.
The proof follows by noticing that every continuous function on a compact set attains its
maximum value.
If the utility function is continuous and quasi-concave, then the solution to the opti-
mization problem is a convex set. If the utility function is in addition strictly quasi-concave,
then the solution is unique.
If the solution is unique, then we can write it as a (multivalued) function of prices and
income p and m. We denote this function by x(p, m) and call it the Marshallian demand
function. If we x prices then x(p, m) is a function of one variable, namely income. This
function is called the Engel function. Engel functions can easily be estimated from cross
sections of household consumption data given that we assume that all households face the
same prices.
If u is locally non-satiated then the budget constraint is binding.
If u(x) is dierentiable and the solution x 0, then it can be found by setting up the
Lagrangian

L(x, ) = u(x) [px m].

The rst order conditions give,

u(x )
pj = 0,
xj
p x = m.

Later on, we will see that we can interpret as the marginal utility of income: how
much more utility can the DM obtain if we marginally increase the amount of income m.
The marginal utility of income gives a conversion between money and utility. So if we
divide the marginal utility of good j by we are measuring the additional amount of
utility that we would get, measured in monetary terms, by increasing xj by one unit.
CHAPTER 1. CONSUMER BEHAVIOUR 11

Theorem 1.4

If u is quasi-concave, strictly monotone and C 1 and (x , ) 0 solve the rst


order conditions, then (x , ) also solve the consumer optimization problem.

step I We rst demonstrate a short intermediate result on quasi-concave functions that if


u is strict quasi-concave and C 1 , then u(x ) u(x) implies u(x) (x x) 0.

Proof. Let u(x ) u(x). By quasi-concavity


u(tx + (1 t)x) min{u(x), u(x )} 0,
u(x + t(x x)) u(x)
0,
t
u(x + t(x x)) u(x)
lim 0,
t0 t
x u(x) (x x) 0.

step II Assume that the rst order conditions hold for (x , ) 0. This means that
x u(x ) = p and px = m. Towards a contradiction, assume that there is another
bundle x X such that px m and u(x ) > u(x ), i.e. x is not optimal.
By strict monotonicity of u, there is a x x such that u(x ) u(x ). Furthermore,
px < px m. Then from quasi-concavity and u(x ) u(x ), we obtain
x u(x ) (x x ) = p (x x ),
= (p x m),
< (m m),
= 0,
a contradiction with quasi-concavity.
Theorem 1.5

Let x 0 solve the consumers maximisation problem at prices p 0 and


income m > 0. If

u is C 2 on Rn++ .
u(x )
xi
> 0 for some i = 1, ..., n, and,

The bordered Hessian of u has a non-zero determinant at x , then x(p, m)


is dierentiable at (p, m).
CHAPTER 1. CONSUMER BEHAVIOUR 12

1.4 Indirect utility and expenditure


1.4.1 The indirect utility function
The optimal value of the utility maximization problem is called the indirect utility
function.

v(p, m) = max u(x) s.t. px m.


x

It gives the maximal utility that can be obtained facing prices p and income m. The
indirect utility function can also be obtained by taking the Marshallian demand function
x(p, m) and substituting this into the utility function: v(p, m) = u(x(p, m)).

Theorem 1.6

If u is continuous and strictly increasing on Rn+ then v(p, m) is

Continuous.

Homogeneous of degree zero in (p, m).

Strictly increasing in m.

Decreasing in p.

Quasi-convex in (p, m).

moreover it satises
v(p,m)

Roys identity v(p,m) = xj (p, m) if v(p, m) is dierentiable.


pj

step I From previous results we know that x(p, m) is continuous. As such, v(p, m) =
u(x(p, m)) is a continuous function of a continuous function, so it is also continuous.

step II For homogeneity, notice that the budget restriction remains the same if both p
and m are multiplied by the same number t > 0, so the utility maximization problem
remains the same and, hence, the optimal value does not change.

step III Let m increase to m . Then we see that px(p, m) < m . As such, there is a bundle
x in an open neighbourhood of x(p, m) such that px m and u(x ) > v(p, m).
This means that v(p, m ) > v(p, m). The intuition is that the constraint px m is
relaxed if m increases. By local non-satiation, the objective function must strictly
increase if m increases.
CHAPTER 1. CONSUMER BEHAVIOUR 13

step IV If p increases, the constraint becomes more stringent. As such, the maximization
problem can not reach a higher value:
step V Finally, for quasi-convexity, consider two prices p, p and two income levels m, m
with optimal consumption bundles x and x . Let p = tp + (1 t)p and m =
tm + (1 t)m . Let x be the optimal solution for (p , m ). We need to show that
v(p , m) max{v(p, m), v(p , m)}. Observe that,

tpx + (1 t)p x = p x m = tm + (1 t)m

As such, either px m or p x m . If the rst is the case, then x was available


when x was chosen, so v(p, m) = u(x) u(x ) = v(p , m ). If the second is the case,
then x was available when x was chosen so v(p , m ) = u(x ) u(x ) = v(p , m )
As such, u(x ) = v(p , m ) max{u(x), u(x )} = max{v(p, m), v(p , m )} which
needed to be shown.
step VI Assume that v(p, m) is C 1 . Then, we can apply the envelope theorem.
v(p, m)
= (p, m) xj (p, m),
pj
v(p, m)
= (p, m).
m
Here (p, m) is the optimal value of the Lagrange multiplier for the utility maxi-
mization problem. From this, we see that (p, m) measures the increase in utility
that can be obtained by marginally increasing the income m. Therefore, it is also
called the marginal utility of income (how much additional utility do you obtain by
increasing the budget by 1 euro).
Taking the ratio gives,
v(p,m)
pj
v(p,m)
= xj (p, m).
m

This shows how to recover the Marshallian demand functions based on the indirect
utility function from the indirect utility functions. It is called Roys identity.

Example:

Consider the indirect utility function v(p, m) = i i ln(pi ) + ln(m). From non-
increasingness in pi we have that i 0. From linear homogeneity, we also have that,

v(tp, tm) = i ln(tpi ) + ln(tm) = i ln(t) + ln(t) + v(p, m).
i i

As such, we must have that i i = 1.
CHAPTER 1. CONSUMER BEHAVIOUR 14

From Roys identity, we have that,


i m
xi (p, m) = .
pi
These demand functions correspond to the Cobb-Douglass utility function.

1.4.2 The Expenditure function


Instead of maximizing utility subject to a linear budget constraint it is also possible to do
the reverse, namely, minimize total expenditure subject to some minimal level of utility
that the consumer would like to attain. We call this the expenditure minimization problem.
Let u0 be some level of utility. The expenditure minimization problem is then,

min ph s.t. u(h) u0 .


hX

X such that u(h)


If u is continuous and there exists a value h u0 , then a solution
to the expenditure minimization problem exists.
If p 0 and u is strictly quasi-concave and strictly monotone, then the solution is
unique.
Proof. If h and h with h = h both provide a solution to the expenditure minimization
problem, then obviously ph = ph . From strict quasi-concavity, we have that u(0.5h +
0.5h ) > u0 so the bundle (0.5h + 0.5h ) is feasible. From strict monotonicity, we can nd a
bundle h 0.5h+0.5h such that u(h ) u0 . However ph < p(0.5h+0.5h ) = ph = ph
which contradicts the assumption that h and h were optimal.
The unique solution to the expenditure minimization problem is denoted by h(p, u0 )
and is called the Hicksian or compensated demand function (the book uses the
notation xc ). Its value determines what the decision maker would demand at prices p to
reach a certain utility level u0 .
The optimal value of the expenditure minimization problem e(p, u0 ) is called the ex-
penditure function.

e(p, u0 ) = min ph s.t. u(h) u0 .


hX

The value of the expenditure function can also be obtained by multiplying the Hicksian
demands by the prices e(p, u0 ) = ph(p, u0 ).
CHAPTER 1. CONSUMER BEHAVIOUR 15

Theorem 1.7

If u(.) is continuous and strictly increasing, then e(p, u0 ) is

Zero when u0 takes on the lowest level of utility.

Continuous,

for all p 0 strictly increasing in u0 and unbounded from above

increasing in p,

Homogeneous of degree one in p,

Concave in p.

If in addition u is strictly quasi-concave then,

Shephards lemma, e(p,u)


pi
= hi (p, u).

Step I If u = u then the entire set X is feasible. Hence also the bundle 0 which gives the
expenditure level 0.

Step II Continuity follows from the fact that e(p, u0 ) = ph(p, u0 ) and h is continuous.

Step III It increases in u0 because the constraint becomes tighter if u0 increases. In order
to see this, assume towards a contradiction that u0 > u0 and that e(p, u0 ) e(p, u0 ).
Then, u(h(p, u0 )) > u0 . As such, by continuity of u(.), there exists a bundle h
h(p, u0 ) such that u(h ) u0 and ph < e(p, u0 ) e(p, u0 ). However, this means
that e(p, u0 ) is not the minimal value of the expenditure minimization problem, a
contradiction.

Step IV It increases in p because the objective function is increasing in p.

Step V For homogeneity, let p = tp. Then u(h(tp, u0 )) u0 and u(h(p, u0 )) u0 . As


such,

tph(tp, u0 ) tph(p, u0 ),
ph(p, u0 ) ph(tp, u0 ).

This shows that the two inequalities should in fact be equalities. From this, it follows
that,

e(tp, u0 ) = tph(tp, u0 ) = tph(p, u0 ) = te(p, u0 ).


CHAPTER 1. CONSUMER BEHAVIOUR 16

Step VI For concavity let p = tp + (1 t)p and let h, h , h be the optimal solutions
at (p, u0 ), (p , u0 ) and (p , u0 ). All bundles can reach the utility level u0 so they are
all feasible. As the bundles are optimal, we have,

ph = e(p, u0 ) ph ,
p h = e(p , u0 ) p h ,
te(p, u0 ) + (1 t)e(p , u0 ) p h = e(p , u0 ).

Step VII this follows from the envelope theorem. Further, notice that because e(p, u0 ) is
homogeneous of degree 1 in prices, this means that h(p, u0 ) is homogeneous of degree
zero in prices.

1.4.3 Relations between the two


The expenditure and indirect utility function are intimately linked.

Theorem 1.8

Let v(p, m) and e(p, u0 ) be the indirect utility function and expenditure function
for some consumer whose utility function is continuous and strictly increasing.
Then for all p 0, m 0, and u0 ,

v(p, e(p, u0 )) = u0 ,

e(p, v(p, m)) = m.

step I For the rst, let x(p, e(p, u0 )) solve the problem max u(x) s.t. px e(p, u0 ) and
let h(p, u0 ) solve min ph s.t. u(h) u0 .
Then ph(p, u0 ) = e(p, u0 ), so the bundle h(p, u0 ) is feasible for the utility maximiza-
tion problem. This implies that v(p, e(p, m)) = u(x(p, e(p, u0 ))) u(h(p, u0 )) u0 .
Now assume, towards a contradiction, that v(p, e(p, u0 )) > u0 . Then, by def-
inition, u(x(p, e(p, u0 ))) = v(p, e(p, u0 )) > u0 . As such, there exists a bundle
x x(p, e(p, u0 )) such that u(x ) u0 . Given that x is feasible in the expen-
diture minimization problem, we must have that, e(p, u0 ) px . However, this
implies that,

e(p, u0 ) px(p, e(p, u0 )) > px e(p, u0 ).

a contradiction.

step II Let x(p, m) be the optimal solution of max u(x) s.t. px m and let h(p, v(p, m))
be the optimal solution to min ph s.t. u(h) v(p, m).
CHAPTER 1. CONSUMER BEHAVIOUR 17

By denition, we have that u(x(p, m)) = v(p, m) so x(p, m) is feasible in the expen-
diture minimization problem. This shows that,

e(p, v(p, m)) px(p, m) m.

Now assume, towards a contradiction, that e(p, v(p, m)) = ph(p, v(p, m)) < m.
Consider a bundle h h(p, v(p, m)) so that ph < m. By strict monotonicity
u(h ) > u(h(p, v(p, m)) v(p, m).
As h is feasible in the utility maximization problem, we have that,

v(p, m) = u(x(p, m)) u(h ) > u(h(p, v(p, m)) v(p, m),

a contradiction. Conclude that e(p, v(p, m)) = m.

Above theorem shows that e(p, u) and v(p, m) are inverse functions. You can obtain
v(p, m) by setting m = e(p, u) and inverting with respect to u. Similarly, you can obtain
e(p, u) by setting v(p, m) = u and inverting with respect to m.
Example:

Consider the indirect utility function v(p, m) = i i ln(pi ) + ln(m). Then substitut-
ing u0 = v(p, m) and e(p, u0 ) = m gives,

u0 = i ln(pi ) + ln(e(p, u0 )),
i

ln(e(p, u0 ) = u0 + i ln(pi ),
i
i
e(p, u0 ) = eu0 eln(pi ) ,
i

e(p, u0 ) = e u0
pi i .
i

Theorem 1.9

Under Assumption 1.2 we have the following relations between the Hicksian and
Marshallian demand functions for p 0, m 0,

xi (p, m) = hi (p, v(p, m)).

hi (p, u) = xi (p, e(p, u)).

Proof. We will consider a somewhat dierent proof as in the book. Assume that the indirect
utilty function v(p, m) and e(p, u0 ) are dierentiable, we obtain, from the previous theorem
CHAPTER 1. CONSUMER BEHAVIOUR 18

that,

v(p, e(p, u0 )) = u0 ,
v(p, e(p, u0 )) v(p, e(p, u0 )) e(p, u0 )
+ = 0,
pi m pi

v(p,e(p,u))
Dividing by m
and applying Roys identity and Shephards lemma gives,

xi (p, e(p, u0 )) = hi (p, u0 ).

In other words, the Marshallian demand at prices p and expenditure level e(p, u0 ) is equal
to the Hicksian demand at prices p and utility u0 . Also

e(p, v(p, m)) = m,


e(p, v(p, m)) e(p, v(p, m)) v(p, m)
+ = 0,
pi u0 pi
e(p, v(p, m)) v(p, m)
= 1.
u0 m
Substituting gives,

hi (p, v(p, m)) = xi (p, m).

This states that the Hicksian demand at prices p and utility v(p, m) is equal to the Mar-
shallian demands at prices p and income m.

1.5 Properties of Consumer Demand


1.5.1 Relative prices and income
Theorem 1.10

Under Assumption 1.2, the consumer demand function xi (p, m), i = 1, ..., n,
is homogeneous of degree zero in all prices and income, and it satises budget
balancedness, px(p, m) = m for all (p, m).

The proof is easy. If both prices and income are multiplied by the same number, the budget
constraint remains the same and the optimal solution to the utility maximization problem
should therefore also remain the same.
Homogeneity of degree zero means that we can either normalize the price of one of the
goods or the income to one, e.g. xi (p, m) = xi (p/m, 1).
CHAPTER 1. CONSUMER BEHAVIOUR 19

1.5.2 Income and substitution eects


Consider the identity,
xi (p, e(p, u0 )) = hi (p, u0 ).
Take the derivative with respect to pj .
xi (p, e(p, u0 )) xi (p, e(p, u0 )) e(p, u0 ) hi (p, u0 )
+ = ,
pj m pj pj
xi (p, e(p, u0 )) xi (p, e(p, u0 )) hi (p, u0 )
+ hj (p, u0 ) = .
pj m pj
the second equality uses Shephards lemma. Substututing m = e(p, u0 ) and v(p, m) = u0
gives,
Theorem 1.11

xi (p, m) hi (p, v(p, m)) xi (p, m)


= xj (p, m).
pj pj m

This equation is called the Slutsky equation. The left hand side is called the total price
eect of the Marshallian demand. It determines how much the demanded quantity changes
due to a change in one of the prices. On the right hand side, we have the sum of two terms.
The rst is called the substitution eect. It measures how much the demand changes due
to the price change (keeping utility constant). The second term is called the income eect.

Theorem 1.12
hi (p,u0 )
The own substitution eect pi
is negative.

Proof. Observe that by Shephards lemma


2 e(p, u0 ) hi (p, u0 )
= .
pi pi pi
This last term is negative because e(p, u0 ) is concave.

Theorem 1.13

The Law of Demand. A decrease in the own price of a normal good will cause
quantity demanded to increase. If an own price decrease causes a decrease in
quantity demanded, the good must be inferior.
CHAPTER 1. CONSUMER BEHAVIOUR 20

Proof. From the Slutsky equation, we obtain that,


xi (p, m) hi (p, v(p, m)) xi (p, m)
= xi (p, m).
pi pi m
The rst term on the right is negative due to the previous theorem. The second term,
however can be both positive or negative. It is negative if good i is normal: income
increase leads to an increase in demand. In this case, the total price eect will also be
negative and we obtain a downward sloping demand function. The Marshallian demand
function will be steeper than the Hicksian demand.
If the second term is positive this is due to the fact that the good is inferior. An income
increase leads to a decrease in demand. In this case, the Marshallian demand function will
be atter than the Hicksian demand function. If the good is suciently inferior, it can even
be that the total price eect is positive. In this case, we call the good a Gien good.

Theorem 1.14

The cross substitution eect is symmetric.

hi (p, u0 ) hj (p, u0 )
= .
pj pi

Proof. By Youngs theorem.

2 e(p, u0 ) hi (p, u0 ) 2 e(p, u0 ) hj (p, u0 )


= = = .
pi pj pj pj pi pi

Theorem 1.15

The substitution matrix p h(p, u0 ) is negative semidenite.

This follows from the fact that this matrix contains the second order derivatives of the
expenditure function which is concave in prices.

Theorem 1.16

Symmetric and Negative Semidenite Slutsky Matrix

This follows from the previous theorem and the Slutsky equation.
CHAPTER 1. CONSUMER BEHAVIOUR 21

1.5.3 Some elasticity relations


Denition 1.6
xi (p,m) m
The income elasticity is dened as i = m xi (p,m)
. The price elasticity is
xi (p,m) pj
dened as i,j = pj xi (p,m)
.
pi xi (p,m)
The income share si = m
.

The elasticity i,j measures the percentage change in the demanded quantity of good i,
xi due to a one percent change in the price of good j, pj . The elasticity i second measures
the percentage change in the demanded quantity of good i, xi due to a one percent change
in income, m.

Theorem 1.17

Let x(p, m) be the consumers Marshallian demand system. Then the following
relations must hold among income shares, price, and income elasticities of demand:

Engel aggregation i si i = 1.

Cournot aggregation i si i,j = sj .


Proof. Step I If we dierentiate the budget constraint i pi xi (p, m) = m with respect to
m, we obtain,

pi xi (p, m) = m,
i
xi (p, m)
pi = 1,
i
m
pi xi (p, m)
i = 1,
i
m

si i = 1.
i

This shows that the weighted sum of all income elasticities must be equal to one.
This identity is called Engel aggregation. It implies that not all goods can be
inferior (elasticity < 0) and also that not all goods can be luxuries (elasticity > 1)
or necessities (elasticity < 1).
CHAPTER 1. CONSUMER BEHAVIOUR 22

Step II if we dierentiate the budget constraint with respect to pj we obtain,


xi (p, m)
pi + xj = 0,
i
pj
pj xi (p, m)
pi i,j + sj = 0,
m i pj

si i,j = sj .
i

This condition is called Cournot aggregation. If there are only two goods, we
obtain,

s1 1,1 + s2 2,1 = s1 ,
s1 (1 + 1,1 ) + s2 2,1 = 0.

If good 1 is price-elastic (1,1 < 1) then good 1 and 2 must be gross substitutes,
2,1 > 0.

The Marshallian demand function x(p, m) is also homogeneous of degree zero in prices
and income; for all t > 0,
x(tp, tm) = x(p, m).
This follows from the fact that multiplying both prices and income by the same amount
does not change the budget constraint. Dierentiating this identity with respect to t and
evaluating at t = 1 gives,

xi (tp, tm) = xi (p, m),


xi xi
pj + m = 0,
j
p j m

xi pj i,j + xi i = 0,
j

i,j + i = 0.
j

In other words the price elasticities and the income elasticity should add up to zero. Ob-
serve that all these conditions follow from the fact that the budget constraint is binding
(so this is independent of the fact that the consumer is utility maximizing).

1.6 Exercises
See tutorials I and II
Chapter 2

Topics in Consumer Theory

2.1 Duality, a closer look


2.1.1 Expenditure and consumer preferences
Consider any function E(p, u) which may not necessarily be an expenditure function and
assume that E(p, u) satises all assumptions of the expenditure function of theorem 1.7.
We will show that in this case, E must be an expenditure function for some preferences.
We will give an explicit construction of such utility function.
Fix (p0 , u0 ) and consider the value E(p0 , u0 ). Then construct the half space,

A(p0 , u0 ) = {x|p0 x E(p0 , u0 )}.

Keep u0 xed and consider a dierent price p1 , construct,

A(p1 , u0 ) = {x|p1 x E(p1 , u0 )}.

Do this for dierent prices and form the innite intersection,



A(u0 ) = A(p, u0 ).
p

Theorem 2.1

Let E : Rn++ R+ R+ satisfy properties 1 through 7 of an expenditure function


given in Theorem 1.7. Then the function u : Rn+ R given by u(x) max{u
0|x A(u)} is increasing, unbounded above, and quasiconcave.

Proof. Let x x . Then there is a p such that,

px = E(p, u(x)).

23
CHAPTER 2. TOPICS IN CONSUMER THEORY 24

From the denition, we also know that for this p,


px E(p, u(x )).
Then,
E(p, u(x)) = px px E(p, u(x )).
As E is strictly increasing in its second argument, u(x) > u(x ).
For quasi-concavity, let x and x be two bundles. Then there is a price vector p such
that,
p(tx + (1 t)x ) = E(p, u(tx + (1 t)x )).
Further, from the denition,
px E(p, u(x)),
px E(p, u(x )).
As such,
E(p, u(tx + (1 t)x )) = p(tx + (1 t)x ) = tpx + (1 t)px tE(p, u(x)) + (1 t)E(p, u(x )).
As such,
E(p, u(tx + (1 t)x )) min{E(p, u(x)), E(p, u(x ))},
u(tx + (1 t)x ) min{u(x), u(x )}.

Theorem 2.2

Let E(p, u) satisfy properties 1 to 7 of an expenditure function given in Theorem


1.7 and let u(x) be derived from E as in Theorem 2.1. Then for all non-negative
prices and utility, E(p, u) = minx px s.t. u(x) u. That is, E(p, u) is the
expenditure function generated by derived utility u(x).

Proof. Let x = p E(p, u). We show the theorem in several steps. First of all, notice that
by Eulers theorem and linear homogeneity of E(p, u),
E(p, u) = pp E(p, u) = px.
Next let us show that u(x) = u. Indeed, for all p, p , by concavity of E(p, u),
E(p , u) E(p, u) (p p)p E(p, u) = (p p)x = p x E(p, u).
As such, p x E(p , u) for all p and for one p: px = E(p, u). So, by the denition of
u(x), we see that u(x) = u.
Finally, let us shows that x solves the expenditure minimization problem. If not, then
there is a x such that u(x ) u and px < px. But then,
px < px = E(p, u).
This shows that u(x ) < u, a contradiction.
CHAPTER 2. TOPICS IN CONSUMER THEORY 25

2.1.2 Convexity and monotonicity


Result: monotonicity and convexity have no testable implications in the theory of consumer
demand.

2.1.3 Indirect utility and consumer demand


We know that u(x) = max{u|p : px e(p, u)}. Inverting the expenditure function, this
is equivalent to,
u(x) = max{u|p : u v(p, px)}.
In other words,
u(x) = min v(p, px).
p

Theorem 2.3

Suppose that u(x) is quasiconcave and dierentiable on Rn++ with strictly positive
partial derivatives there. Then for all x Rn++ , v(p, px), the indirect utility
function generated by u(x), achieves a minimum in p on Rn++ , and

u(x) = min v(p, px).


p

The proof is demonstrated above. Notice that the rst order conditions give,
v(p, px) v(p, px)
+ xi = 0,
pi m
v(p,px)
pi
xi = v(p,px)
.
m

This is Roys identity.


The solution to the minimization of the expenditure function is called the indirect
demand function p(x). It gives the prices at which the bundle x is demanded. In general,
this price vector will not be unique because if p is a solution, then tp will also be a solution.
As such, we should normalize the prices. Usually, this is done by setting px = 1.
Theorem 2.4

Let u(x) be the consumers direct utility function. Then the inverse demand func-
u(x)
pi (x) xi
tion for good i associated with income y = 1 is given by p(x)x
= u(x) . This
i xi xi
is called Wolds theorem. If we normalize prices we can drop the denominator.
CHAPTER 2. TOPICS IN CONSUMER THEORY 26

By the envelope theorem, we have that

u(x) v(p, px)


= pi (x),
xi m
u(x) v(p, px) v(p, px)
xi = pi (x)xi = m.
i
x i m i
m

Taking the ratio gives,


u(x)
pi (x) xi
= u(x)
.
p(x)x i xi
xi

This shows how to recover the inverse demand function from the direct utility function. It
is called Wolds theorem.

2.2 Integrability
In reality, we do not observe the utility or expenditure functions. However, it is possible
to obtain information or estimate the Marshallian demand funtions x(p, m).
The question that we would like to answer is what conditions need to be satises on such
demand function x(p, m) such that there exists a utility function that generates x(p, m)
as the outcome of the utility maximization problem? This is called the integerability
problem.
In turns out that three requirements must be satised:

budget balancedness px(p, m) = m.

homogeneity of degree zero x(tp, tm) = x(p, m).

Slutsky conditions The Slutsky matrix is symmetric and negative semi denite.

In fact, the rst and third condition imply the second.

Theorem 2.5

If x(p, m) satises budget balancedness and its Slutsky matrix is symmetric, then
it is homogeneous of degree zero in p and m.

Proof. For a proof let fi (t) = xi (tp, tm) we need to show that fi (t) is a constant function.

fi (t) xi (tp, tm) xi (tp, tm)


= pj + m.
t i
pj m
CHAPTER 2. TOPICS IN CONSUMER THEORY 27

From budget balance i tpi xi (tp, tm) = tm so m = i xi (tp, m)pi . Therefore,

fi (t) xi (tp, tm) xi (tp, tm)


= pj + xj (tp, tm)pj ,
t j
pj m j
[ xi (tp, tm) xi (tp, tm) ]
= + xj pj .
j
p j m

The term in brackets is the compensated price eect, so we may swith i and j.
[ ]
fi (t) xj (tp, tm) xj (tp, tm)
= + xi pj ,
t j
pi m
xj (tp, tm) xj (tp, tm)
= pj + xi pj ,
j
pi j
m
1 xj (tp, tm) 1 xj (tp, tm)
= tpj + xi tpj ,
t j pi t j m
1 1
= (xi (tp, tm)) + xi (tp, tm) ,
t t
= 0.

So we have that a demand function should satisfy, budget balancedness, Slutsky nega-
tivity and Slutsky symmetry.

Theorem 2.6

A continuously dierentiable function x : Rn+1++ R is the demand function


n

generated by some increasing, quasiconcave utility function if (and only if, when
utility is continuous, strictly increasing, and strictly quasiconcave) it satises
budget balancedness, symmetry, and negative semideniteness.

The proof is in two steps. In step 1, it is shown that we can construct a function e(p, u)
that satises all properties of an expenditure function. In step 2 this function is used to
construct the utility function u(x). In fact, we already demonstrated step 2, so we only
need to focus on step 1.
Consider some demand function x(p, m). Assume that there exist a function e(p, u)
such that the following dierential equations are satised.

e(p, u)
= xi (p, e(p, u)).
pi
CHAPTER 2. TOPICS IN CONSUMER THEORY 28

Let us rst show that such function satises all properties of an expenditure function.
Concavity is satised by the fact that the Slutsky matrix (which is the matrix of second
order partial derivatives of e(p, u) is negative semi denite). Also, by concavity of e(p, u),

e(tp, u) e(p, u) p e(p, u)(tp p),


= (t 1)px(p, e(p, u)) = (t 1)e(p, u).

This shows that e(tp, u) te(p, u). Similarly,

e(p, u) e(tp, u) p e(tp, u)(p tp),


(1 t)
= (1 t)px(tp, e(tp, u) = e(tp, u).
t
As such, te(p, u) e(tp, u). This shows that e(tp, u) = te(p, u), i.e. the function e(p, u) is
homogeneous of degree 1. The function e(p, u) is increasing in p because the rst derivative
is non-negative (i.e. xi (p, e(p, u)) 0).
In order to complete the proof we still need to show that above dierential equation
has a solution for e(p, u). This is established by something that is called the Frobenius
theorem. This theorem says that a solution to a dierential equation exists if and only if
the cross partial derivatives are equal. Or,

xi (p, e(p, u)) xi (p, e(p, u) e(p, u) xj (p, e(p, u)) xj (p, e(p, u) e(p, u)
+ = +
pj m pj pi m pi
xi (p, e(p, u)) xi (p, e(p, u) xj (p, e(p, u)) xj (p, e(p, u)
+ xj (p, e(p, u)) = + xi (p, e(p, u))
pj m pi m

However, this boils down to the condition that the Slutsky matrix is symmetric.

2.3 Revealed preferences


The integrability problem described above assumed that we know the complete demand
function. However, in reality, what we normally observe is at best a nite number of prices
p1 , p2 , . . . , and a nite number of corresponding consumption bundles x1 , x2 , . . .. Let us
gather these in a nite data set S = {pt , xt }tT . Here t is called the set of observations.
What can we now say in terms of the underlying utility function? This is the question
answered by the theory of revealed preferences.
Assume that we have observed two price vectors pt , pv and the associated consumed
bundles xt , xv . If these bundles were generated by utility maximizing behaviour (that
satises local non-satiation), then the budget must be exhausted so we can reconstruct the
incomes mt = pt xt and mv = pv xv .
What can we say about the preferences based on the observations (pt , xt ) and (pv , xv )?
Well, if mt = pt xt pt xv , then xv was feasible at the budget Bt = {x|pt x mt }.
However, we observe that xt was chosen.
CHAPTER 2. TOPICS IN CONSUMER THEORY 29

Then, it must then be that u(x1 ) u(x2 ). If pt xt pt xv we say that xt is directly


revealed preferred to xv and we write xt RD xv . The binary relation RD is called the
direct revealed preference relation.
If utility is strictly quasi-concave and monotone, then demand is unique. As such, if
xt RD xv and xt = xv , then u(xt ) > u(xv ).
Now, assume that xt = xv . If xt RD xv and xv RD xt (i.e. pv xv pv xt ) we obtain
that u(xt ) > u(xv ) and u(xv ) > u(xt ), so these observations cannot be generated by a
utility maximizing decision maker. This shows that the constructed relation RD should be
asymmetric (xt RD xv implies not xv RD xt if xt = xv ). This condition is called the Weak
Axiom of Revealed Preference or WARP.

Denition 2.1

A consumers choice behaviour satises WARP if for every distinct pair of bundles
xt , xv with xt chosen at prices pt and xv chosen at prices pv , pt xt pt xv
pv xv < pv xt .

WARP and budget balancedness implies homogeneity of degree zero. WARP also
implies Slutsky negativity. In order to see this, consider a price pt and consumed bundle
xt . Assume that prices change, but let us adjust the budget so that mv = pv xt , which
means that the bundle xt can still be bought at this new situation. Assume that for
this new budget and prices, the DM chooses bundle xv . By construction pv xv = pv xv ,
so xv RD xt . Then, if WARP is satised, we should have that pt xt < pt xv . Taking the
dierence of these two inequalities gives,

(pt pv )xt < (pt pv )xv ,


(pt pv )(xt xv ) < 0.,

the income compensated price eect is negative. It can be shown that this is equivalent
to negativity of the Slutsky matrix so WARP is equivalent to Slutsky negativity.
Let pv = pt + kz, then,

(kz) [x(pt , mt ) x(pv , pv xt )] < 0.,


(kz) [x(pt , mt ) x(pt + kz, mt + kzxt )] < 0,
zx(pt , mt ) > zx(pt + kz, mt + kzxt ).

The function on the right hand side is maximized for k = 0. As such, its derivative with
respect to k should be non-positive at k = 0. Then,
[ xi (pt , mt ) xi (pt , mt ) ]
zi + xj (pt , mt ) zj 0.
i j
pj m
CHAPTER 2. TOPICS IN CONSUMER THEORY 30

The term between brackets is the substitution eect (i.e. the elements of the Slutsky
matrix). Given that z was arbitrary, this condition states that the Slutsky matrix should
be negative semi-denite.
However, this says nothing about symmetry of the Slutsky matrix. It turns out that
WARP is not sucient to impose symmetry (except if there are only two goods). In this
case, a stronger condition is necessary, called Strong Axiom of Revealed Preference
or SARP.
SARP states that for all nite sequences of distinct bundles (pt , xt ), t = 1, . . . , T
xt RD xk , xk RD xr , . . . , xz RD xv then pv xv < pv xt . Equivalently, SARP requires that the
direct revealed preference relation RD is acyclic. SARP is obviously necessary for utility
maximizing behaviour. Indeed, a violation of SARP would imply that, u(xt ) > u(xk ) >
u(xr ) > . . . > u(xv ) > u(xt ). Moreover, it turns out that it is also sucient. However the
proof of this is rather lengthy and complicated so we will omit it here.
What happens if we drop strict quasi-concavity. In this case, demand can be mutli-
valued. The condition in this case is called the Generalized Axiom of Revealed Pref-
erence or GARP: if xt RD xk , xk RD xr , . . . , xz RD xv then pv xv pv xt . GARP replaces
the strict inequality in the closing condition with a weak inequality. GARP is necessary
and sucient for consistency with locally nonsatiated utility function and even consistency
with a strict monotone and concave utility function.

2.4 Uncertainty
Until now, we assumed that everything is perfectly known to the DM. Real life, however,
is full of uncertainties. In this part, we look at how a decision maker deals with problems
in the face of uncertainty.

2.4.1 Preferences
Consider a nite set of outcomes X = {a1 , . . . , an }.

Denition 2.2
[ ]
A simple gamble is a vector of numbers p = p1 p2 . . . , pn such that out-
come a1 occurs with probability p1 , outcome a2 with probability p2 , . . . and out-
come an occurs with probability pn . Of course, we need to impose that pi 0
(nothing can happen with negative probability) and i pi = 1 which means that
at least something should happen.

Note, in this section we use p for a vector of probabilities not for a vector of prices.
CHAPTER 2. TOPICS IN CONSUMER THEORY 31

We denote a simple gamble also as (p1 a1 , . . . , pn an ) and denote by G0 the set of all
simple gambles.

G0 = {(p1 a1 , . . . , pn an )|ai X, pi = 1, pi 0}.
i

Given the set of simple gambles, it becomes possible to construct gambles of simple gambles,
e.g. if g1 and g2 G0 then (p g1 , (1 p) g2 ) describes the gamble where g1 occurs with
probability p and g2 with probability g2 . We can denote this by the set G2 . We can iterate
this and denote the set Gn as the set of gambles of gambles of Gn1 . The set of all gambles
G is then the union of all Gi over n N.
Every gamble g can be reduced to a simple gamble by computing out the nal probabil-
ities of the events a1 , . . . , an . This gamble is called the reduced gamble of g and is denoted
by gs .
For example, if g1 = (p a1 , (1 p) a2 ) and g2 = (q1 a1 , (1 q) a2 ) then the reduced
gamble of the gamble ( g1 , (1 ) g2 ) is given by ((p + (1 )q) a1 , ((1 p) + (1
)(1 q)) a2 )
The preference relation R is dened on the set of all gambles G (R G G) that
satises the standard axioms.

Axiom 1 Completeness: for any two gambles, g and g in G, either gRg or g Rg.

Axiom 2 Transitivity: for any three gambles g, g , g in G, if gRg and g Rg , then gRg .

Axiom 3 Continuity: the preference relation is continuous if for any gamble g in G, there
is some probability, [0, 1], such gI(a1 , (1 )an ).

Consider the set of outcomes X. the gamble (0 a1 , 0 a2 , . . . , 1 ai , 0 ai+1 , . . .) is in G


so the preference relation R also imposes a ranking on all outcomes in X. Let us re-rank
all outcomes in X such that a1 Ra2 Ra3 , . . . , Ran . We impose several additional conditions
on R.

Axiom 4 The preference relation R is said to be monotone if for all , [0, 1]

( a1 , (1 ) an )R( a1 , (1 ) an ),

if and only if .
If the probability of the most desirable outcome increases and the probability of the
least desirable outcome decreases, then the decision maker is not worse o.
CHAPTER 2. TOPICS IN CONSUMER THEORY 32

Axiom 5 The preference relation R satises substitution if for all g = (p1 g1 , . . . , pn


gn ) and h = (p1 h1 , . . . , pn hn ) then, if gi Ihi for all i, then gIh.

Axiom 6 The preference relation R satises reduction to simple gambles if gIgs


where gs is the reduced gamble of g.

This axiom requires that the decision maker is smart enough to compute the reduced
gamble gs from any compound gamble g and notice that they induce the same lottery over
nal outcomes.
The combination of axioms 5 and 6 form the key axioms and are not innocuous. It as-
sumes that we can exchange gambles in composed gambles by indierent gambles. Observe
that it also implies that,
gIh ( g, (1 ) h)Ig.
Having a mixture between two indierent lotteries is equivalent to having one of the two
lotteries.

2.4.2 Von Neumann-Morgenster utility


If preferences are complete transitive and continuous, then there exist a continuous utility
function such that u(g) u(g ) if and only if gRg . However by adding the other axioms,
one can obtain a much more stringent functional form on u.

Denition 2.3

A utility function u : G Rn has the expected utility property if for all g G,



u(g) = pi u(ai ),
i

where gs = (p1 a1 , . . . , pn an ) is the simple gamble induced by g.

Observe what happens here. First the utility is computed for each outcome ai giving
the values u(ai ). Next the expected value of these numbers is computed. The existence of
a utility function with the expected utility property is guaranteed by the following result.

Theorem 2.7

If the preferences R satises Axioms 1-6 then it can be represented by a utility


function with the expected utility property.
CHAPTER 2. TOPICS IN CONSUMER THEORY 33

Proof. As before, let a1 be the best outcome and an the worst, and dene the numbers
u(g) such that,

gI(u(g) a1 , (1 u(g)) an ).

We claim that u(g) has the expected utility form.


First of all, let us show that u(g) is well dened. Let W (g) = {t|gR(t a1 , (1 t) an )
and let B(g) = {t|(t a1 , (1 t) an )Rg}. Then W (g) and B(g) are closed by continuity
and W (g) is of the form [r, 0] and B(g) is of the form [k, 1]. By completeness, the union
of these two sets is [0, 1] so r k. By monotonicity they are equal. So u(g) is indeed a
function.
Let gs be the simple gamble induced by g. Then, by reduction to simple lotteries, gs Ig
so,

(u(g) a1 , (1 u(g)) an )I(u(gs ) a1 , (1 u(gs )) an ).

by monotonicity, it follows that u(g) = u(gs ). Let gi be the lottery such that,

gi = (u(ai ) a1 , (1 u(ai )) an ).

Then applying substitution twice and reduction to simple lotteries, we obtain,

gs I(p1 g1 , . . . , pn gn )I(p1 a1 , (1 p1 ) an )

Where p1 is given by,



p1 = pi u(ai ).
i

As such, by transitivity,

gs I( pi u(ai ) a1 , (1 pi u(ai )) an )
i i

By monotonicity, we have that u(gs ) = i pi u(ai ).
The following result shows that the expected utility representation is unique up to a
linear transformation.
Theorem 2.8

Suppose that the VNM utility function u() represents R. Then the VNM utility
function, v(), represents those same preferences if and only if for some scalar
and some scalar ,

v(g) = + u(g).
CHAPTER 2. TOPICS IN CONSUMER THEORY 34

Proof. Suciency is easy because u and g give the same ranking over all gambles. For
necessity assume that u and v give the same ranking. Let
ai I(i a1 , (1 i ) an ).
We know that u(ai ) = i u(a1 ) + (1 i )u(an ) and v(ai ) = i v(a1 ) + (1 i )v(an ). Then,
u(ai ) u(an ) v(ai ) v(an )
= i = .
u(a1 ) u(an ) v(a1 ) v(an )
So,
(u(ai ) u(an ))(v(a1 ) v(an )) = (v(ai ) v(an ))(u(a1 ) u(an )).
(v(ai ) v(an ))
u(ai ) = (u(a1 ) u(an )) + u(an ),
(v(a1 ) v(an ))
u(a1 ) u(an ) v(an )
u(ai ) = v(ai ) (u(a1 ) u(an )) + u(an ),
v(a1 ) v(an ) (v(a1 ) v(an ))
u(ai ) = v(ai ) + .
and are independent of i so they are the same for all ai . further > 0.
Then,

u(g) = pi u(ai ) = pi v(ai ) + pi ,
i i i

= pi v(ai ) + b,
i
= av(g) + b.

2.4.3 Risk aversion



Consider the case where X = R and ai are monetary outcomes. Let E(g) = i pi ai be the
expected value of the gamble.
Denition 2.4

Let u() be an individuals VNM utility function for gambles over non-negative
levels of wealth. Then for the simple gamble g = (p1 w1 , . . . , pn wn ), the
individual is said to be

risk averse at g if u(E(g)) > u(g),

risk neutral at g if u(E(g)) = u(g),

risk loving at g if u(E(g)) < u(g).


CHAPTER 2. TOPICS IN CONSUMER THEORY 35

Denition 2.5

The certainty equivalence of the gamble g is the outcome C(g) such that,

u(C(g)) = u(g).

The certainty equivalence computes what the individual wants with certainty to be indif-
ferent to the gamble.
The individual is risk averse if and only if C(g) < E(g). In order to see this, notice
that, if the individual is risk averse, then

u(E(g)) > u(g) = u(C(g)).

The proof follows from monotonicity of u. On the other hand, if C(g) < E(g), we have,

u(g) = u(C(g)) < u(E(g)).

The risk premium is dened as the dierence between the expected value of the
gamble and the certainty equivalence, r(g) = E(g) C(g). It is the maximal amount of
money that the individual would be willing to give up order to substitute the gamble for
the expected value of the gamble.

Denition 2.6

The Arrow-Pratt measure of absolute risk aversion is given by,

u (w)
Ra (w) = .
u (w)

Consider two individuals with utility functions v and u. We can always dene a function
h such that for all w R, v(w) = h(u(w)). Furthermore, h will be strictly increasing by
monotonicity of the utility function.
Then,

v (w) = h (u(w))u (w),


v (w) = h (u(w))u (w)u (w) + h (u(w))u (w),
v (w) h (u(w))u (w)u (w) h (u(w))u (w)
= ,
v (w) h (u(w))u (w) h (u(w))u (w)
( )
v (w) u (w) u (w)
= h (u(w)) .
v (w) u (w) h (u(w))
CHAPTER 2. TOPICS IN CONSUMER THEORY 36

The ratio u (w)/h (u(w)) is positive so the dierence in the measure of absolute risk aversion
is positive if h (u(w)) < 0 or equivalently, if h is concave. As such, an individual will have
a higher absolute risk aversion for all levels of welfare if his utility function is a concave
transformation of the utility function of the other individual.
Now consider a gamble g and consider the certainty equivalents Cv (g) and Cu (g) of the
two individuals and let v(w) = h(u(w)) where h is a concave function. Then,

v(Cv (g)) = pi h(u(ai )),
(
i
)

<h pi u(ai ) ,
i
= h(u(g)),
= h(u(Cu (g))) = v(Cu (g))

This shows that the certainty equivalence of for the utility function u is higher than the
certainty equivalence of v if h is concave. In other words, the more risk averse individual
has a lower certainty equivalence.
The Arrow-Pratt measure of relative risk aversion is given by,

u (w)
Ra (w) = w.
u (w)
Chapter 3

Theory of the rm

3.1 Primitive notions


Remember that a decision making problem was dened as a combination of a choice space,
a choice set and a preference relation or utility function.
In a production context, the choice space of the rm is the space of all possible input
output combinations. The choice set corresponds to the input output combinations that
can be produced with a certain technology. The preference relation is replaced with a
production function. Let us make this a bit more specic.

3.2 Production
A production possibility set Y is a subset of Rn . This represents set of all possible input-
output combinations. Usually, one takes the convention that negative amounts in the vector
are inputs and positive amounts are outputs. This is also called a netput. The production
possibility set is a subset of X which contains all netputs that are technologically feasible.
We assume that
0 Y : it is possible for the rm to produce nothing.

if y Y , and y Rn+ then y = 0: it is impossible to produce something without


inputs.

Y is closed and bounded from above.


In many cases, we will consider the setting where the rm produces only one output,
using n inputs. This gives us a netput vector of (y, x1 , . . . , xn ). We denote the inputs
also as x = (x1 , . . . , xn ). In such setting, we dene a production function as,

f (x) = max y s.t. (y, x) Y


y

f (x) gives the maximal output that can be produced using the input vector x.

37
CHAPTER 3. THEORY OF THE FIRM 38

Assumption 3.1 The production function, f : Rn+ R+ is continuous, strictly increas-


ing, and strictly quasiconcave on Rn+ and f (0) = 0.
The production function gives a ranking on the set of inputs, where xRx if and only
if f (x) f (x ). The production isoquant is then the set I(x) which gives the combi-
nations of inputs that generate the same output as f (x). Similarly to the marginal rate of
substitution in a production context, we dene the marginal rate of technical substitution
as,
f (x)
xi M Pi (x)
M RT Si,j = f (x)
= .
M Pj (x)
xj

f (x)
This gives the slope at the production isoquant. The function xj
is called the marginal
productivity of input j and is also written as M Pi (x).

Denition 3.1

Let N = 1, ..., n index the set of all inputs, and suppose that these inputs can be
partitioned into S > 1 mutually exclusive and exhaustive subsets, N1 , ..., NS . The
production function is called weakly separable if the M RT S between two inputs
within the same group is independent of inputs used in other groups:

fi (x)/fj (x)
= 0,
xk
for all i, j Ns and k N Ns .

Denition 3.2

For a production function f (x), the elasticity of (substitution of input


) j for input
i at the point x0 R++ is dened as i,j (x0 ) = ln(M RT Si,j (x(r))) where x(r) is
n ln(r)

the unique vector of inputs x = (x1 , ..., xn ) such that

(i) xj /xi = r,

(ii) xk = xk,0 for k = i, j, and

(iii) f (x) = f (x0 ).

The elasticity of substitution measures the curvature of the production isoquant. The elas-
ticity is big if the curvature is nearly linear: in this case both goods are very substitutable.
The elasticity is small if the production isoquant is very bended. In such settings the inputs
are not very substitutable.
CHAPTER 3. THEORY OF THE FIRM 39

Theorem 3.1

Let f (x) be a production function satisfying Assumption 3.1 and suppose it is


homogeneous of degree ]0, 1]. Then f (x) is a concave function of x.

Proof. Assume that f is homogeneous of degree 1 and let y1 = f (x1 ) and y2 = f (x2 ) then,
( ) ( )
x1 x2
f =f = 1.
y1 y2
Because f is strictly quasi-concave, we have that,
( )
x1 x2
f t + (1 t) 1.
y1 y2
Picking t = y1 /(y1 + y2 ) gives,
( )
x1 + x2
f 1,
y1 + y2
f (x1 + x2 ) y1 + y2 = f (x1 ) + f (x2 ).
Let x1 = tx and x2 = (1 t)x , then,
f (tx + (1 t)x ) f (tx) + f ((1 t)x ) = tf (x) + (1 t)f (x ).
This shows that f is concave if it is homogeneous of degree 1. Finally, if f is homogeneous
of degree < 1, then f 1/ is homogeneous of degree 1 and is therefore concave. As f is a
concave transformation of a concave function, it is also concave.

3.2.1 Returns to scale and varying proportions


Denition 3.3

A production function f (x) has the property of (globally):

Constant returns to scale if f (tx) = tf (x) for all t > 0 and all x;

Increasing returns to scale if f (tx) > tf (x) for all t > 1 and all x;

Decreasing returns to scale if f (tx) < tf (x) for all t > 1 and all x.

Constant returns to scale means that if we increase the inputs by a factor of t, then output
also increases by a factor t. In other words, if the production function is homogeneous of
degree one. In this case, taking the derivatives with respect to t and setting t = 1 gives,

M Pi (x)xi = f (x).
i
CHAPTER 3. THEORY OF THE FIRM 40

Denition 3.4

The scale elasticity is given by,



f (tx) t i M Pi (x)xi
lim = .
t1 t f (tx) f (x)

So the scale elasticity is the sum of the output elasticities.

It is easy to see that constant returns to scale implies a scale elasticity equal to one.
Increasing returns to scale implies a scale elasticity larger than one and decreasing returns
to scale implies a scale elasticity smaller than one.

3.3 Cost
In the consumer model, we saw the expenditure minimization model. In a production
context, the corresponding model is given by the cost minimization model. It gives the
minimal cost necessary for the rm to reach a certain level of output.
Denote the input prices by the vector w. The minimal costs of producing a certain
amount of output y is then determined by,
min wx s.t. f (x) y.
x

Observe the similarity with the cost minimization problem in consumer choice. The solu-
tion of this problem are called the conditional input demands and are denoted by x(w, y).
The optimal value of the problem is called the cost function and is denoted by c(w, y).

Denition 3.5

The cost function, dened for all input prices w 0 and all output levels y is
the minimum-value function,

c(w, y) min wx s.t. f (x) y.


x

If x(w, y) solves the cost-minimisation problem, then c(w, y) = wx(w, y).

Similar as for the consumer problem, we can set up the Lagrangian and derive the rst
order conditions, these give,
f (x)
wi = ,
xi
f (x) = y.
CHAPTER 3. THEORY OF THE FIRM 41

Taking the ratio of the rst condition for two inputs gives that the ratio of input prices
is equal to the MRTS, i.e. the slope of the iso-cost line is equal to the slope at the iso-
production isoquant.

Theorem 3.2

If f is continuous and strictly increasing, then c(w, y) is

Zero when y = 0,

Continuous on its domain,

For all w 0, strictly increasing and unbounded above in y,

Increasing in w,

Homogeneous of degree one in w,

Concave in w.

Moreover, if f is strictly quasiconcave we have Shephards lemma:

c(w, y)
= xi (w, y).
wi
i = 1, ..., n.

The proof is the same as for the expenditure minimization problem.

Theorem 3.3

Suppose the production function satises Assumption 3.1 and that the associated
cost function is twice continuously dierentiable. Then

x(w, y) is homogeneous of degree zero in w,

The substitution matrix, dened and denoted


x1 (w,y)
w1
. . . xw
1 (w,y)

.. ,
n

(w, y) = ... ..
. .
xn (w,y) xn (w,y)
w1
... wn

is symmetric and negative semidenite. In particular, the negative semidef-


initeness property implies that xw
i (w,y)
i
0 for all i
CHAPTER 3. THEORY OF THE FIRM 42

Theorem 3.4

When the production function satises Assumption 3.1 and is homothetic,

(a) the cost function is multiplicatively separable in input prices and out-
put and can be written c(w, y) = h(y)c(w, 1), where h(y) is strictly
increasing and c(w, 1) is the unit cost function, or the cost of 1 unit of
output;
(b) the conditional input demands are multiplicatively separable in input
prices and output and can be written x(w, y) = h(y)x(w, 1), where
h (y) > 0 and x(w, 1) is the conditional input demand for 1 unit of
output.

When the production function is homogeneous of degree > 0,

(a) c(w, y) = y 1/ c(w, 1);


(b) x(w, y) = y 1/ x(w, 1).

Proof. The production function is homothetic if f (x) = g(r(x)) for some h homogeneous
of degree 1. Let h(.) = g 1 (.). Then we have that

min wx s.t. f (x) y


= min wx s.t. g(r(x))/y = 1,
( ( ))
x
= min wxs.t. g r = 1,
h(y)
= min h(y)wz s.t. f (z) = 1,
= h(y)c(w, 1).

Then, Shephards lemma states that xi (w, y) = h(y) c(w,1)


wi
= h(y)xi (w, 1).
CHAPTER 3. THEORY OF THE FIRM 43

Denition 3.6

Let the production function be f (z), where z (x, x ). Suppose that x is a


subvector of variable inputs and x is a subvector of xed inputs. Let w and w
be the associated input prices for the variable and xed inputs, respectively. The
short-run, or restricted, total cost function is dened as

yx
sc(w, w, ) = min wx + w ) y.
x s.t. f (x, x

If x(w, w, y; x
) solves this minimisation problem, then sc(w, w, y; x) =

wx(w, w, y; x) + w
x. The optimised cost of the variable inputs, wx(w, w, y; x ),
x, is called total xed
is called total variable cost. The cost of the xed inputs, w
cost.

It is clear from the denition that long run costs can never be bigger than short run
costs.
Let x (w; w,
y) be the optimal value of xed inputs that minimize short run costs.

(w, w,
x y) = arg min sc(w, w,
y; x
).
x

The rst order condition gives,

y; x
sc(w, w, )
= 0.
xi
y; x
We know that sc(w, w, (w, w,
y)) = c(w, w,
y). As such, we have that,
sc(w, w,
y; x) xi (w, w,
y) y; x
sc(w, w, ) y)
c(w, w,
+ = ,
i
xi y y y
y; x
sc(w, w, (w, w,
y)) y)
c(w, w,
= ,
y y

This shows that the short run marginal costs (evaluated at the optimal amount of xed
inputs x) is equal to the long term marginal cost. If two functions take the same value
at some point and have the same slope, they are tangent. As such, the long term cost
function is the lower envelope of the short run cost functions.

3.4 Duality in production


Obtaining the production function from the cost function is identical to the consumption
framework.
CHAPTER 3. THEORY OF THE FIRM 44

Theorem 3.5

Let c : Rn++ R+ R+ satisfy properties 1 to 7 of a cost function given in


Theorem 3.2. Then the function f : Rn+ R+ dened by

f (x) max{y 0|wx c(w, y), w 0}

is increasing, unbounded above, and quasiconcave. Moreover, the cost function


generated by f is c.

Theorem 3.6

A continuously dierentiable function x(w, y) mapping Rn++ R+ Rn+ is the


conditional input demand function generated by some strictly increasing, quasi-
concave production function if and only if it is homogeneous of degree zero in w,
its substitution matrix, whose ijth entry is xwi (w,y)
j
, is symmetric and negative
semidenite, and wx(w, y) is strictly increasing in y.

3.5 The competitive rm


Prefect competition states that the rm is price taker on the output and input market.
The idea is that there is erce competition. If a rm asks a price above the market
price, it will sell nothing. If it asks a price below the market price, it can make larger
prots by increasing the price a little bit. As such, it will always choose to sell at the price
prevailing in the market.

3.5.1 Prot maximisation


Let p be the output price. The prots of the rm are given by,

max py wx s.t. f (x) y.


x

We can replace the inequality by an equality and obtain the problem maxx pf (x) wx.
The rst order condition requires that,

f (x)
p = wi .
xi
The left term is called the marginal revenue of an input. The right hand is the marginal
cost of one additional input.
CHAPTER 3. THEORY OF THE FIRM 45

Instead of maximizing prots, one could also think of a two step process. First costs
are minimized for given levels of output, giving the cost function c(w, y). Next, y is chosen
optimally in order to maximize prots.

max py c(w, y).


y

c(w,y)
This gives the condition p = y
or price equal to marginal cost.

3.5.2 The prot function


The solution to the prot maximization problem gives the input or factor demand
functions x(p, w) and the output supply function y(p, w) = f (x(p, w)) or simply the
supply function. The prot function is the optimal value (p, w) = py(p, w) wx(p, w).

Denition 3.7

The rms prot function depends only on input and output prices and is dened
as the maximum-value function,

(p, w) = max py wx s.t. f (x) y.


(x,y)

This requires that the optimum of the prot function exists. This question is a bit more
involved that it sounds.
The production maximization problem is not well dened if the production function
has increasing or constant returns to scale. In that case, we have that by multiplying the
inputs by t > 1,

pf (tx) wtx t(pf (x) wx) t(p, w).

This shows that (if prots are positive), then under non-decreasing returns to scale, the
prot function will be unbounded and the problem is not well dened.
CHAPTER 3. THEORY OF THE FIRM 46

Theorem 3.7

If f satises Assumption 3.1, then for p 0 and w 0, the prot function


(p, w), where well-dened, is continuous and

Increasing in p,

Decreasing in w,

Homogeneous of degree one in (p, w),

Convex in (p, w),

Dierentiable in (p, w) 0. Moreover, under the additional assumption


that f is strictly concave (Hotellings lemma),

(p, w) (p, w)
= y(p, w), = xi (p, w).
p wi

Proof. The single less obvious property is convexity. Let p, p , w, w be two output input
prices and let p = tp + (1 t)p , w = tw + (1 t)w . Let (x, y), (x , y ) and (x , y ) be
the corresponding solutions to the prot maximization problem.
Then we know that,

py wx py wx ,
py wx py wx .

As such,

t(py wx) + (1 t)(p y w x ) p y w x ,


t(p, w) + (1 t)(p , w ) (p , w ).

This also implies that if prices are uncertain, then the expected prot of the rm will be
larger than the prots obtained from the expected prices.
CHAPTER 3. THEORY OF THE FIRM 47

Theorem 3.8

Suppose that f is a strictly concave production function satisfying Assumption 3.1


and that its associated prot function, (p, w), is twice continuously dierentiable.
Then, for all p > 0 and w 0 where it is well dened:

Homogeneity of degree zero: y(tp, tw) = y(p, w) for all t > 0, xi (tp, tw) =
xi (p, w) for all t > 0 and i = 1, ..., n.

Own-price eects:

y(p, w) xi (p, w)
0, 0,
p wi
for all i = 1, ..., n.

The substitution matrix


y(p,w) y(p,w) y(p,w)

...
xp w1 wn

1p(p,w)
xw
1 (p,w)
. . . xw
1 (p,w)

1 n ,
.. .. .. ..
. . . .
p
xn (p,w)
w1
xn (p,w)
... xn (p,w)
wn

is symmetric and positive semidenite.

Analogue to the short run cost function, there is also a short run prot function
CHAPTER 3. THEORY OF THE FIRM 48

Theorem 3.9

Suppose that f : Rn+ R+ is strictly concave and satises Assumption 3.1.


For k < n, let x Rk+ be a subvector of xed inputs and consider f (x, x ) as
a function of the subvector of variable inputs x R+ . Let w and w
nk
be the
associated input prices for variable and xed inputs, respectively. The short-run,
or restricted, prot function is dened as

x
(p, w, w; ) max py wx w ) y.
x s.t. f (x, x
y,x

x
The solutions y(p, w, w; ) and x(p, w, w,
x ) are called the short-run, or restricted,
output supply and variable input demand functions, respectively. For all p > 0
and w 0, (p, w, w, x ) where well-dened is continuous in p and w, increasing
in p, decreasing in w, and convex in (p, w). If (p, w, w; x
) is twice continuously
x
dierentiable, y(p, w, w, ) and x(p, w, w,
x ) possess all three properties listed in
Theorem 5.8 with respect to output and variable input prices.

Observe that,

x
(p, w, w, ) = max py sc(w, w,
y; x
).
y

As such,

y; x
sc(w, w, )
p= .
y
In order for the rm to decide whether to produce or not in the short run she must
compare the prots when producing with the prots when producing nothing.

w
x (p, w, w,
x ).

This requires that the price should be above the average variable cost.

3.6 Exercises
Chapter 5

General Equilibrium

In previous sections, we derived the optimal behaviour for one single consumer and one sin-
gle rm. In reality, however, there are many rms and many consumers who simultaneously
decide how much to consume and produce. Furthermore, markets are interrelated.
The total demanded quantities and supplied quantities are a function of the prices faced
in the markets. A market is said to be in equilibrium if for every good, total demanded
quantity is equal to the total supplied quantity.
A price vector is said to be an equilibrium price vector if all the markets are in
equilibrium simultaneously. Several issues are worth studying.
Does there exist an equilibrium?
What are the welfare properties of such an equilibrium?
Is the equilibrium unique?
Is the equilibrium stable?
We will focus on the rst two questions.

5.1 Equilibrium in Exchange


Let us start with a simple setup without production. Assume that there are n individuals
where individual i has a utility function ui and an initial vector of endowments ei R+ .
Here is the amount of goods. If i consumes the bundle xi she obtains utility ui (xi ).
In an exchange economy, it is assumed that total endowment is xed e = i ei .
Individuals start with their own endowments but are allowed to exchange goods with each
other on a voluntary basis. So goods will be exchanged if both individuals prot from this
exchange.
i A feasible allocation is a vector of consumption bundles (x1 , . . . , xn ) such that
i x = e. In other words, the total amount of endowments are redistributed among the
agents in the economy. We denote by F (e) the set of feasible allocations.

F (e) = {xi , i n| xi = e}.
i

49
CHAPTER 5. GENERAL EQUILIBRIUM 50

Denition 5.1

A feasible allocation, x F (e), is Pareto ecient if there is no other feasible


allocation, y F (e), such that yi Ri xi for all consumers, i, with at least one
preference strict.

Pareto eciency embeds the notion of unanimity.

Denition 5.2

Let S I denote a coalition of consumers. We say that S blocks x F (e) if


there is an allocation y such that

i
iS y = iS e
i

yi Ri xi for all i S, with at least one preference strict.

The intuition is that the individuals in S can reallocate their own resources among them-
selves in order to obtain a better outcome. Notice that the allocation is Pareto optimal if
and only if the grand coalition N = {1, . . . , n} is not blocking.

Denition 5.3

The core of an exchange economy with endowment e, denoted C(e), is the set of
all unblocked feasible allocations.

Given that no single individual must be blocking and the grand coalition is also not blocking
a minimal requirement is that C(e) contains only Pareto optimal allocations where ui (xi )
ui (ei ). This second condition is called individual rationality. In a two agent economy,
Pareto optimality and individual rationality characterize the core. If exchange is voluntary,
it would be expected that nal allocations end up in the core of the economy.

5.2 Equilibrium in competitive market systems


5.2.1 Existence of equilibrium
We impose the following assumption on the utility functions of the individuals.
CHAPTER 5. GENERAL EQUILIBRIUM 51

Assumption 5.1 Utility ui is continuous, strongly increasing, and strictly quasiconcave


on Rn+ . On competitive markets, every consumer takes prices as given, whether acting as a
buyer or a seller. If p (p1 , . . . , p ) 0 is the vector of market prices, then each consumer
solves

max
i
ui (xi ) s.t. pxi pei .
x

Theorem 5.1

If ui satises Assumption 5.1 then for each p 0, the consumers problem has a
unique solution, xi (p, pei ). In addition, xi (p, pei ) is continuous in p on Rn++ .

Denition 5.4

The aggregate excess demand function for good k is the real-valued function,

zk (p) = xik (p, pei ) eik .
iN iN

The aggregate excess demand function is the vector-valued function,

z(p) [z1 (p), ..., zn (p)] .

In general the excess demand function will depend on the distribution of the endowments
so in principle, we should write, z(p, e1 , . . . , en ). In many cases, however we use z(p) under
the assumption that the distribution of endowments is xed. The excess demand function
for a certain good is positive if total demand for this good is larger than total supply, it
is below zero if supply is larger than demand. We say that the market for good j is in
equilibrium if zj (p) = 0. We say that the price vector p is an equilibrium price vector if
all markets are simultaneously in equilibrium.

Denition 5.5

A vector p Rn++ is called a Walrasian equilibrium if z(p ) = 0.

The equilibrium price vector should be such that in each market, aggregate demand is
equal to aggregate supply.
CHAPTER 5. GENERAL EQUILIBRIUM 52

Theorem 5.2

If for each consumer i, ui satises Assumption 5.1, then for all p 0,

Continuity: z(.) is continuous at p.

Homogeneity: z(p) = z(p) for all > 0.

Walras law: p z(p) = 0.

Homogeneity implies that we can normalize the price of one of the goods to one. Walras
law also implies that,

pj zj (p) = pi zi (p).
j=i

So if we are at an equilibrium on n 1 markets (e.g. if zj (p) is zero for all j = i), then it
must be the case that zi (p) is also equal to zero. In other words, if n 1 markets are at
equilibrium then the last market should also be in equilibrium. Now notice that,

if z(p) 0 and p 0, then z(p) = 0.

If not, then we would have that pz(p) < 0 which would contradict Walras law.

Theorem 5.3

Suppose z : Rn++ Rn satises the following three conditions:

z(.) is continuous on Rn++ ;

pz(p) = 0 for all p 0.

if {pm } is a sequence of price vectors in Rn++ converging to p = 0 and pk = 0


for some good k, then for some good k with pk = 0, the associated sequence
of excess demands in the market for good k , {zk (pm )}, is unbounded above.

Then there is a price vector p 0 such that z(p ) = 0.

The third condition roughly states that if the price for some goods converge to zero,
then the
excess demand for at least one of those goods should go to innity. If pj 0
then i
i iN xi k (p, pei ) for some k with pk 0. However, in this case zk (p) =
i xk (p, pe ) i ek which means that market k is not in equilibrium. As such,
i

condition 3 guarantees that in a Walrasian equilibrium, all prices should be strictly positive.
CHAPTER 5. GENERAL EQUILIBRIUM 53

In order to show the existence of a Walrasian equilibrium, we use Brouwers xed


point theorem

Any continuous mapping from a nonempty compact set A Rn to A, has a


fixed point, i.e. an element p A such that f (p) = p.

Observe that by Walras law, we only need to show that there exists a p such that z(p) 0
and p 0.
Using homogeneity, we normalize the prices such that j pj = 1. This denes the
domain of the mapping to the n 1 dimensional simplex n1
{ }

n1
= p| pj = 1 ,
j

which is a nonempty and compact set.


Consider the functions
fj (p) = pj + zj (p).
Now if fj (p) = pj for all j, we have that pj = pj + zj (p) which implies that zj (p) = 0 and
we would be done. Unfortunately, we cannot apply Brouwers xed point theorem to this
function because the range of the function f (p) is not the n 1 dimensional simplex.
One way to map into the n 1 dimensional simplex would be to the normalize the
function such that its elements sum to one,

pj + zj (p) pj + zj (p)
fj (p) = = .
k pk + k zk (p) 1 + k zk (p)

Now, if the price of some goods converge to zero, we saw that zj (p) converged to innity

for some goods. This would give us the value fj (p) = , which poses a problem. In order
to avoid this, we need to make sure that the denominator and numerator remain bounded.
One way to do this is by specifying the function rj (p) = min{zj (p), 1},

pj + rj (p)
fj (p) = .
1 + k rk (p)

The idea is that we are in fact only interested in points where zj (p) = 0 so bounding this
function from above at 1 does not impose any loss of information. We also need to resolve
the potential problem that the denominator may be zero, which is again undesirable.
Remember that we actually only need to show that for some p 0 zj (p) 0 for all
j. This means that we can put,

pj + max{rj (p), 0}
fj (p) = .
1 + k max{rk (p), 0}
CHAPTER 5. GENERAL EQUILIBRIUM 54

Now, the denominator is strictly positive. However, the additional requirement is now
that p 0 which means that fj (p) should map to the interior of the simplex. We do
this by adding a small number a to the numerator (and to normalize, we add na to the
denominator). Then,

pj + a + max{rj (p), 0} a a
fj (p) = > 0.
na + 1 + k max{rk (p), 0} na + 1 + n 2n + 1

Let n1 (a) be the subset of the n 1 dimensional simplex such that { i pi = 1, p
a/(1 + 2n)}. This is the n 1 dimensional simplex excluding a tiny section of its boundary.
So we know that f (p) maps into n1 (a). Now, consider a xed point of f (p).

a + pj + max{rj (p ), 0}
pj
= ,
na + 1 + k max{rl (p ), 0}
( )

pj na + 1 + max{rk (p ), 0} = a + pj + max{rj (p ), 0},
k

pj (na + max{rk (p ), 0}) = a + max{rj (p ), 0},
k

Now, we let a 0. Let us show hat p converges to some strict positive vector. If not,
there is some good i for which pi 0. However, by assumption, there should be some
other good m for which pm 0 and zm (p) . For this good, the left hand side becomes
zero while the right hand side converges to 1, which is impossible. As such, the sequence
of prices will converge to some price vector p 0 and satisfy the condition,

pj max{rk (p ), 0} = max{rj (p ), 0},
k

pj zj (p ) max{rk (p ), 0} = zj (p ) max{rj (p ), 0},
k

max{rk (p ), 0} pj zj (p ) = zj (p ) max{rj (p ), 0}.
k j j

The left hand side is (by Walras law) zero. The terms zj (p ) max{rj (p ), 0} on the right
hand side cannot be strict negative. Because if zj (p ) < 0 then the product is zero and if
zj (p ) > 0 then the product is strict positive. The sum of non-negative terms can only be
zero if every term is zero. This means that every term on the right hand side is zero. This
implies that either zj (p ) = 0 or zj (p ) < 0. In other words, zj (p) 0 for all j, which
establishes what we needed to show.
If the demand functions of the individuals are a correspondence instead of a function,
then we need an extended xed point theorem.
CHAPTER 5. GENERAL EQUILIBRIUM 55

Theorem 5.4

If each consumers utility function satises Assumption


5.1, and if the aggregate
endowment of each good is strictly positive (i.e., i ei 0), then aggregate
excess demand satises conditions 1 through 3 of Theorem 5.3.

Theorem 5.5

If each consumers utility function satises Assumption 5.1, and iI ei 0,
then there exists at least one price vector, p 0, such that z(p ) = 0.

5.2.2 Eciency
We have shown that an exchange equilibrium with volutary exchange will intuitively lead
to an allocation in the Core. Next we have see that there exist a Walrasian equilibrium.
In this part, we have a look at the connection between the two.

Denition 5.6

Let p be a Walrasian equilibrium for some economy with initial endowments


e, and let x(p ) (x1 (p , p e1 ), ..., xI (p , p eI )), where component i gives the
n-vector of goods demanded and received by consumer i at prices p . Then x(p )
is called a Walrasian equilibrium allocation, or WEA.

Lemma 5.1

Let p be a Walrasian equilibrium for some economy with initial endowments e.


Let x(p ) be the associated WEA. Then x(p ) F (e).

This follows immediately from the denition of a Walrasian equilibrium.


Lemma 5.2

Suppose that ui is strictly increasing on Rn+ , that consumer is demand is well-


dened at p 0 and equal to x i , and that xi Rn+

If ui (xi ) > ui (
xi ), then pxi > p
xi .

If ui (xi ) ui (
xi ), then pxi p
xi .
CHAPTER 5. GENERAL EQUILIBRIUM 56

Proof is by simple revealed preference argument.

Denition 5.7

For any economy with endowments e, let W (e) denote the set of Walrasian equi-
librium allocations.

Theorem 5.6

Consider an exchange economy (ui , ei ), i I. If each consumers utility function,


ui , is strictly increasing on Rn+ , then every Walrasian equilibrium allocation is in
the core. That is,

W (e) C(e).

Proof. Assume, towards a contradiction, that it is not. Let (p, x) be a Walrasian allocation
and consider a blocking coalition S. This means that there is an allocation z for the
members ofS, such that ui (zi ) ui (xi ) for all i S with at least one strict inequality and

iS e . From Lemma 5.2 above, we have that for all i S,


i i
iS z =

pzi pxi ,

With
at ileast one strict inequality. However, this implies
that p iS zi > p iS xi =
i i
p iS e . However, this contradicts the fact that iS z = iS e . This gives the
desired contradiction.
The following is a simple corollary of the theorem above.

Theorem 5.7 (First fundamental theorem of welfare economics)

Under the hypotheses of Theorem 5.6, every Walrasian equilibrium allocation is


Pareto ecient.

Theorem 5.8 (Second fundamental theorem of welfare economics)



Consider an exchange economy (ui , ei ), i I, with aggregate endowment i ei
0, and with each utility function ui satisfying Assumption 5.1. Suppose that x is a
Pareto-ecient allocation for (u , e ), i I, and that endowments are redistributed
i i

so that the new endowment vector is x . Then x is a Walrasian equilibrium


allocation of the resulting exchange economy (ui , x i ), i I.
CHAPTER 5. GENERAL EQUILIBRIUM 57

Proof. Consider a Pareto optimal allocation x. The idea is to implement a so-called no-
trade equilibrium. Set the endowments such that for all i N , ei = xi . Now, if y is
a Walrasian equilibrium forthese endowments,
i then it must be the case that ui (yi )
i i i i i i
u (e ) = u (x ) for all i and i y = i e = i x . However, because x is Pareto optimal,
it must actually be that ui (yi ) = ui (xi ), which implies, by uniqueness of the demand that
y i = xi .
In fact, the redistribution should only be such that the new endowment is on the equilibrium
budget hyperplane for all individuals.

5.3 Equilibrium in production


5.3.1 Producers
Consider a set of rms J , with closed, strongly convex and bounded production possibility
sets Y j containing 0.

Assumption 5.2 The Individual Firm


1. 0 Y j Rn
2. Y j is closed and bounded.
3. Y j is strongly convex. That is, for all distinct y1 , y2 Y j and all t ]0, 1], there
Y j such that y
exists y ty1 + (1 t)y2 and equality does not hold
Every rm maximizes prots,

j (p) = maxj pyj .


yY

The solution gives the supply function yj (p). Aggregate supply is given by j yj (p).

Theorem 5.9

If Y j satises conditions 1 through 3 of Assumption 5.2, then for every price


p 0, the solution to the rms problem (5.3) is unique and denoted by yj (p).
Moreover, yj (p) is continuous on Rn++ . In addition, j (p) is well-dened and
continuous on Rn+ .


The aggregate production possibility set is given by Y = {y = j yj , yj Y j }. It is
what all rms can jointly produce. Let

(p) = max py.


yY
CHAPTER 5. GENERAL EQUILIBRIUM 58

This denes an aggregate supply function y(p) corresponding to a representative rm


maximizing the sum of all individual prots.

Theorem 5.10

If y(p) maximizes joint prots then it is the sum of yj (p)where for each j J,
yj (p) maximizes the individual prots. Formally, y(p) = j yj (p).

Proof. We have,

max py s.t. y Y = max py s.t. y = y j , yj Y j ,
y y
j

= max p yj s.t. yj Y j ,
j

= max pyj s.t. yj Y j .
j

Above result states that the aggregate supply can be seen as the supply of a representative
rm. In other words, supply aggregates nicely. Observe that this is not true for demand.
The aggregate demand function is not necessarily a demand function for a representative
individual.

Theorem 5.11

If each Y j satises Assumption 5.2, then the aggregate production possibility set,
Y , also satises Assumption 5.2.

Theorem 5.12 (Aggregate Prot Maximisation)

For any prices p 0, we have p y py for all y Y if and only if for some
j Y j , j J, we may write y
y = jy yj pyj for all yj , j J.
j and p

5.3.2 Consumers
The prots of the rms need to ow back
to the consumers. We assume that consumer i
i,j i,j
has a share of in rm j. Of course i = 1 for all rms j (shares sum to unity).
CHAPTER 5. GENERAL EQUILIBRIUM 59

The budget restriction for consumer i is then,



px pei + i,j j (p) = mi (p).
j


As before we dene the aggregate demand by i xi (p, mi (p)).

Theorem 5.13

If each Y j satises Assumption 5.2 and if ui satises Assumption 5.1, then a


solution to the consumers problem (5.5) exists and is unique for all p 0.
Denoting it by xi (p, mi (p)), we have furthermore that xi (p, mi (p)) is continuous
in p on Rn++ . In addition, mi (p) is continuous on Rn+ .

5.3.3 Equilibrium

The aggregate supply is now j yj (p) + i ei . Equilibrium price vector p is found where,

z(p ) = xi (p , mi (p )) yj (p ) ei = 0.
i j i

Theorem 5.14

Consider the economy (ui , ei , ij , Y j ), i I, j J.


If each ui satises Assumption
j
5.1, each Y satises Assumption
j 5.2, and y + i ei 0 for some aggregate
production vector y j Y , then there exists at least one price vector p 0
such that z(p ) = 0.

5.3.4 Welfare
Denition 5.8

Let p 0 be a Walrasian equilibrium for the economy (ui , ei , ij , Y j ), i I, j


J. Then the pair (x(p ), y(p )) is a Walrasian equilibrium allocation (WEA)
where x(p ) denotes the vector, (x1 , x2 , . . . , xI ), whose ith entry is the utility-
maximising bundle demanded by consumer i facing prices p and income mi (p );
and where y(p ) denotes the vector, (y1 , y2 , . . . , yj ), of prot-maximising produc-
tion
vectors p . (Note then that because p is a Walrasian equilibrium,
at prices
i i j
iI x = iI e + jJ y ).
CHAPTER 5. GENERAL EQUILIBRIUM 60

Denition 5.9

The feasible allocation (x, y) is Pareto ecient if there is no other feasible al-
location (
x, y xi ) ui (xi ) for all i I with at least one strict
) such that ui (
inequality.

Theorem 5.15 (First Welfare Theorem with Production)

If each ui is strictly increasing on Rn+ , then every Walrasian equilibrium allocation


is Pareto ecient.

Proof. Assume that (x, y) is a WEA but not Pareto ecient. Given that it is a WEA, we
have that,

xi = yj + ei .
i j i

If x is not Pareto ecient, there exist another allocation ( ) such that x = e + y and
x, y
for all i I,

xi ) ui (xi ).
ui (

with at least one strict inequality. By a simple RP argument, this means that,

xi pxi ,
p

with at least one strict inequality. But then



xi >
p pxi ,
i i

p
y +j
ei > pyj + ei ,
j i j i

yj >
p pyj ,
j j

p
y > py j .

This contradicts the fact that yj was prot maximizing.


CHAPTER 5. GENERAL EQUILIBRIUM 61

Theorem 5.16

Suppose that (i) each ui satises Assumption 5.1, (ii) each Y j satises Assumption
5.2, (iii) y + i ei 0 for some aggregate production vector y, and (iv) the
allocation ( ) is Pareto ecient. Then there are income transfers, T 1 , . . . , T I ,
x, y
satisfying i T i = 0, and a price vector, p , such that

i maximizes ui (xi ) s.t. pxi mi (


1. x p) + T i , i I.

yj s.t. yj Y j , j J.
j maximizes p
2. y

Proof. Let Y j = Y j {yj }. This set is also a production technology set. Next, let the
endowment of individual i be equal to x i . Then this economy has a Walrasian equilibrium,
say ( ). Given that 0 Y j prots are non-negative and so every individual can purchase
x, y
his/her own endowment. As such,
xi ) ui (
ui ( xi ).
xi , yi ) was also feasible for the initial economy where yj = y
Let us now show that ( j + y
j
and notice that yj Y j by denition of Y j . Then, given that ( ) is a Walrasian
x, y
equilibrium,

i =
x xi + yj ,
i i j

= xi + (yj y
i ),
i j

i
= e + yj .
i j

As such, we see that ( xi , yj ) was feasible in the initial economy. As such, by Pareto
i all inequalities should be an equality. Next, optimality of x
optimality of x i and strict
quasi-concavity implies that x i = x
i.

5.4 Contingent plans


What happens if there is uncertainty or if we include time; The idea here is to make
commodities state and time contingent. The denition of a good depends on when and in
which state it is consumed.

5.4.1 Time
The idea is here to make commodities time contingent bundles. Preferences are intertem-
poral.
CHAPTER 5. GENERAL EQUILIBRIUM 62

5.4.2 Uncertainty
The idea is here to make bundles state contingent.

5.4.3 Walrasian equilibrium with contingent commodities


There are N goods, T dates and St states per date t. A bundle is a vector consisting of
numbers xkts given the amount of good k at date t in state s. Preferences and production
is over these contingent commodities.
Individuals have endowments which are vectors of numbers eikts . Equilibrium requires
that demand equals supply at every date in every state of the world. Equilibrium prices
specify the price of goods for a given data in a specic state of the world.
The equilibrium condition is that,
j
xikts = ykts + eikts ,
i j i

for all k, t, s. Each individual has a budget constraint,


j

pkts xikts = ij pkts ykts + pkts eikts .
kts j kts kts

This sums over all states. In other words, ex post, it might be that the budget constraint
is not binding. Then how is this allocation implemented? You have to look at it as selling
and buying contracts.

5.5 Core and equilibria


We have seen that every Walrasian equilibrium is in the core. The reverse is not true
though. For a given endowment, there are (in general) innitely many points in the core
that are not a Walrasian equilibrium.
For very large economies, however, there is a one to one correspondence between the
core an a Walrasian equilibrium. The way this is shown is through the use of a replica
economy.

5.5.1 Replica economies


An economy for a nite number of individuals I can be summarized by their utility func-
tions and the endowments E = {u1 , e1 ; u2 , e2 ; . . . ; un , en }. A 2-fold replication of such
economy is an economy of 2n agents where each second agent is an exact copy of an agent
in the original economy. This gives an economy,

E 2 = {u1 , e1 ; u1 , e1 ; u2 , e2 ; u2 , e2 ; . . . ; un , en ; un , en }.

We can generalize this in the following way,


CHAPTER 5. GENERAL EQUILIBRIUM 63

Denition 5.10

Let there be I types of consumers in the basic exchange economy and index these
types by the set I = 1, ..., I. By the r-fold replica economy, denoted Er , we mean
the economy with r consumers of each type for a total of rI consumers. For any
type i I, all r consumers of that type share the common preferences Ri on
Rn+ and have identical endowments ei 0. We further assume for i I that
preferences Ri can be represented by a utility function ui satisfying Assumption
5.1.

For an allocation x in the replica economy, we denote by xiq the allocation for the qth
replication of individual i.
Theorem 5.17

Equal Treatment in the Core. If x is an allocation in the core of Er , then every


consumer of type i must have the same bundle according to x. That is, for every

i = 1, ..., I, xiq = xiq for every q, q = 1, ..., r.

Proof. For each type of individual i I, take (one of) the worst of individual(s) of the type.
Renumber the individuals such that the rst replication
1 is this worst of individual for
each type i. Consider an allocation which gives q xiq /r to this worst of individual of type
i. For at least one type, the worst of type will be strictly better o after this reallocation
because he/she obtains a convex combination of allocations which are (i) not worst than
his/her own allocation and (ii) one of which is distinct from his/her own allocation. Let
i,1 . Then,
us call this allocation x
xiq
i1
=
x
iI iI q
r
1 iq
= x ,
r iI q
1 iq
= e ,
r iI q
1 i1
= re ,
r iI

= ei1 .
iI

This means that this allocation can be implemented by the coalition of worst of individuals.
This, in turn means that this coalition is a blocking coalition. As such, the allocation x
cannot be in the core.
CHAPTER 5. GENERAL EQUILIBRIUM 64

Above theorem shows that agents of the same type should be treated identically.
[ 1 As such,]
a core allocation x in a replica economy can be represented by a vector x , . . . , x , . . . , xI
i

where xi is the allocation given to all agents of type i. Furthermore, by above theorem,
we see that this allocation must also be feasible for the original economy. We know that
every Walrasian equilibrium is a core equilibrium, so every Walrasian equilibrium should
also have the same allocation for all agents of the same type.
Lemma 5.3

The sequence of sets C1 , C2 , . . . containing the core allocations for dierent replica-
economies is decreasing.

Proof. Suppose that x Cr . Now, any coalition that blocks the r 1 fold replica economy
can also block the r-fold replica economy.
Lemma 5.4

An allocation x is a WEA for Er if and only if it is of the form


x = x1 , x1 , . . . , x1 , x2 , x2 , . . . , x2 , . . . , xI , xI , . . . , xI
| {z } | {z } | {z }
r times r times r times

and the allocation (x1 , x2 , ..., xI ) is a WEA for E1 .

Proof. Given that two agents of the same type have the same utility function, they also
i i iq iq
have the same
demand functions
i x (p, pe ) = x (p, pe ).
i i
Then if i x (p, pe ) = i e , we have that,

xiq (p, peiq ) = xiq (p, peiq ) = xi (p, pei ) = ei = eiq .
iI q q iI q iI q iI iI q

This shows that the price vector p is also an equilibrium price vector of the replica economy.
On the other hand, if p is an equilibrium price vector of the replica economy, then

r ei = eiq = xiq (p, peiq ) = xi (p, pei ) = r xi (p, pei ).
i i q iI q iI q iI

so it is also an equilibrium price vector of the base economy.


Theorem 5.18

If x Cr for every r = 1, 2, ..., then x is a Walrasian equilibrium allocation for


E1 .
CHAPTER 5. GENERAL EQUILIBRIUM 65

Proof. Let x be a Pareto optimal allocation of the economy, then there exist welfare
weights i such that,

i i i
x = arg max
i
i u (y ) s.t. y = ei .
y
i i i

The rst order conditions of this optimization program give,

i xi ui (xi ) = p.

Here p is the vector of Lagrange multipliers. Observe that p is independent of i. Setting


i = 1/i this reduces to the well known rst order conditions for the utility maximization
problem.

xi ui (xi ) = i p.

As such, we have the following partial result that if x is a Pareto optimal allocation of the
economy and pxi pei for all i, then x is also a Walrasian equilibrium allocation.
Next if x is in the core, it is Pareto optimal. As such, by the theorem above, if we can
show that for all i, pxi = pei , we are done. Consider the allocation x i = txi + (1 t)ei .
We rst show that ui ( xi ) ui (xi ). Assume towards a contradiction that ui ( xi ) > ui (xi ).
By strict quasi-concavity, this implies that ui (t xi + (1 t )ei ) > ui (xi ) for all t ]t, 1]. By
continuity, there is an integer r such that,
( ( ) )
1 i1
u x 1
i i
+e > ui (xi ).
r r

Now consider the coalition where we have r 1 copies of all types other than i and all r
agents of type i. Assume that ( we give
) all types j = i the allocation xj and the r agents of
type i the allocation zi = xi 1 1r + ei 1r . Then
( )
1 1
(r 1) x + r(x 1
j i
+ ei ),
j=i
r r

=(r 1) xj + rxi xi + ei ,
j=i

=(r 1) xj + ei = (r 1) ei + ei .
j j

As such, this allocation is feasible for this subcoalition, but all agents of type i strictly
improve which means that xi was not in the core of this r-replica economy.
Conclude that

ui (txi + (1 t)ei ) ui (xi ).


CHAPTER 5. GENERAL EQUILIBRIUM 66

Observe that ui (txi + (1 t)ei ) ui (xi ) is negative for all t and zero for t = 1 so the slope
of this function at t = 1 should be positive,

x ui (xi )(xi ei ) 0.

Let p be the vector of Lagrange multipliers for the Pareto optimization program. Then we
have that, such that x ui (xi ) = p/i . Substituting gives,

pxi pei .

Summing over all i gives,



pxi p ei = p xi .
i i i

This shows that all inequalities are in fact equalities. Conclude that,

pxi = pei .

This shows that for large enough economies all Walrasian equilibria will be in the core.

Vous aimerez peut-être aussi