Consumption

Berardino Cesi

University of Rome Tor Vergata, Msc in Economics 2014-2015

Housekeeping 2

I Syllabus
I (Neo-)classical demand theory
I (Neo-)classical production theory
I Choice under uncertainty
I General Equilibrium
I Main textbook:
I Mas-Colell, Whiston & Green, Microeconomic Theory (chaps
1-6)
I Jehle, J A and P J Reny, Advanced Microeconomic Theory
I Useful references are:
I Deaton, A and J Muellbauer, Economics and consumer
behaviour
I Varian Hal, Microeconomic analysis
I Miller, Nolan H., Notes on Microeconomic Theory (available
from his website at Harvard University)
Consumption Theory
Main assumptions 1

Assumptions
I 1 representative consumer
I L commodities, physically dierent from each other
I commodities are available in commodities bundles xs, where
0 1
x1
B ::: C
B C
B x` C
x =B B :::
C
C
B C
@ xL A
Main assumptions 2

I commodity bundles belong to the consumption set X , the set

of consumption bundles which are physically feasible. We will
normally use X = <L+
I consumers have preference relation over commodity
bundles, where preference relation means is at least as well
liked as (or, in short, is at least as good as). Well also use
which means is liked as (or, in short, is as good as)
I consumption set is convex: any linear combination of any two
bundles belonging to the consumption set also belongs to the
consumption set
To x ideas: convex/not convex set

Given a set B RN ; the convex hull of B, denoted Co B, is the

smallest convex set containing B, that is, the intersection of all
convex sets that contain B.
To x ideas: convex/not convex set

Separating Hyperplane Theorem: if B is convex and closed and

x2= B there exists a hyperplane separating point x and set B.

A convex set: a set with no notches and no holes in the convex hull
Properties of the preference relation

Preference rlt.
For all x, y and z 2 X , the preference relation satises:
1. Rationality:
I completeness: either x y or y x
I transitiveness: if x y and y z, then x z
2. Desiderability
I monotonicity : x y implies x y (where x y means
x` > y` for all `) [Strong monotonicity : x` y` for all ` and
x` > y` for at least one ` implies x y ].
Main idea: more is better!!
I local non-satiation: for every " > 0, there is y such that
kx y k " and y x
Local non-satiation

when X = RL+ strong monotonicity implies local non-satiation

although local non-satiation may not imply monotonicity
Consumption set

From these assumptions, for any consumption bundle x, can dene

3 sets of consumption bundles
I indierence set. All consumption bundles that are indierent
to x: fy 2 X : y xg
I upper contour set. All consumption bundles that are at least
as good as x: fy 2 X : y xg
I lower contour set. All consumption bundles that x is at least
as good as: fy 2 X : x y g
Indierence sets
Properties of the preference relation (more)

For all x, y and z 2 X

3. Convexity:
I convexity : A preference relation is convex whenever its
upper contour set is convex.
For any 2 [0; 1] and for any y and z such that y x and
z x, then y + (1 )z x [ y + (1 )z x for strict
convexity]
Main idea: preference for diversication (or, in jargon,
diminishing marginal rate of substitution, see below)
Strictly convex/not convex preference relations
Convex/not convex preference relations
From preference relation to utility function 1

Utility fct
I Useful to represent the preference relation by means of a
utility function.
I A further assumption on the preference relation is needed.
Then, for all x and y 2 X
4. Continuity:
I continuity : A preference relation is continuous if
- the upper contour set fy 2 X : y x g and
- the lower contour set fy 2 X : x y g
are both closed, that is they contain their boundaries.
From preference relation to utility function 2

I Rationality and continuity assumptions are su cient for the

existence of a continuous utility function u(x) that represents

I u(x) needs not to be unique: any z(x) = f (u(x)) where f (:)

is a strictly increasing function, also represents .
I Notice that, of course, continuity does not imply
dierentiability.
I Normally, it is assumed that u(x) is twice dierentiable
From restrictions on to properties of u(x)

Properties of the preference relation translate into properties of

the utility function u(x).
I monotonicity of ) u(:) is increasing:
x y ) x y ) u(x) > u(y )
I convexity of ) quasi-concavity of u(:):
almost tautological:
quasi-concavity of u(x) means that,
for all x,
the set fy 2 <L+ : u(y ) u(x)g is convex.
Quasiconcavity
Quasiconcavity (more)
Quasiconcavity (more)
To x ideas: concave function
To x ideas: quasi-concave function

quasi-concave functions
have
convex upper contours
To x ideas: quasi-convex function

quasi-convex functions
have
convex lower contours
Types of preferences

Types of preferences
Widely used types of preference relations / utility functions are
I Homothetic preferences
I Quasi-linear preferences
I Leontief preferences
Homothetic preferences

I Preferences are homothetic if

all indierence curves are related
by a proportional expansion along rays
I That is, for any x; y such that x y , then x y for any
0.
I Homothetic preference relations are representable by a utility
function that is homogenous of degree one
u( x) = u(x) for all > 0.
Homothetic preferences
Quasi-linear preferences

Preference relation on X = ( 1; 1) <L+ 1

is quasi-linear w. r.
to commodity 1 (the numeraire) if
I commodity 1 is desirable
I indierence set are parallel along the axis of commodity 1
Quasi-linear preference relations are representable by an utility
function u(x) = x1 + (x2 ; : : : ; xL ).
Quasi-linear preferences
Leontief preferences

I A Leontief preference relation is such that x 0 x 00 i

minfx10 ; : : : ; xn0 g minfx100 ; : : : ; xn00 g
I Notice: u(x) is continous but not dierentiable
Leontief preferences
The consumers decision problem

UMP
I Let p 0 and w > 0, with p being commoditiesprices and
w consumers wealth/income
I The consumers decision problem may be described as

max u(x) s.t.:p x w UMP

x 0

I In words:
choose the preferred consumption bundle
within the set of admissible bundles
Budget set 1
Admissible bundles are those within the budget set Bp;w

Bp;w = fx 2 <+ : p x wg
Budget set 2

Important assumptions on prices

I prices p are positive: p 0
I prices p are known
I prices are given to the consumer
I prices are linear: ) budget set Bp;w is convex.
Important assumptions on income/wealth
I income/wealth p is given
Budget set and prices
A budget set which is not convex

s =price
UMP

From UMP, two interesting objects:

I optimal consumption bundles: the solution to UMP
I consumers maximal utility value: the value function of the
UMP
UMP

walrasian demand
From UMP, two interesting objects:
,! optimal consumption bundles: the solution to UMP
I consumers maximal utility value: the value function of the
UMP
The solution to UMP

The solution to the UMP is the Walrasian (or ordinary or market

or Marshallian) demand correspondence (or function): x(p; w )
a rule that assigns optimal consumption vector(s)
to each price-wealth combination
If x(p; w ) is single valued, then Walrasian demand function,
otherwise Walrasian demand correspondence
The solution to UMP
Properties of Walrasian demand correspondence

1. homogeneity of degree zero in p and w :

x(p; w ) = x( p; w ) for any > 0
2. Walrass law
w p x(p; w )
3. x(p; w ) is convex
I if is convex (i.e. u(x) is quasi-concave), then x(p; w ) is a
convex set
I if is strictly convex (i.e. u(x) is strictly quasi-concave), then
x(p; w ) is a single element
Convexity of Walrasian demand
Necessary conditions for solution to UMP

If x 2 x(p; w ) is a solution to UMP, then there exists a

Lagrangean multiplier such that, for all `,
@u(x )
@x p`
`
@u(x )
@x` p` x` = 0 FOC-UMP
In matrix notation
ru(x ) p x (ru(x ) p) = 0
I If u(:) is quasiconcave, FOC-UMP are su cient for global
maximum.
FOC-UMP
Interior solution
Marginal rate of substitution

FOC-UMP imply that, for any `; k

@u(x )
@x` p`
@u(x )
=
pk
@xk |{z}
| {z } price ratio
MRS `;k

I MRS`;k tells how much you want, at the margin, to trade

good ` for good k and keep utility constant.
I If above equality not satised, trading commodities ` and k at
current prices increases utility, contradicting the maximality of
x
Corner solution
MRS at corner solution
Comparative statics

I Useful to look at how demand x(p; w ) changes when w or p

change
I Well look at two objects:
I price/wealth eects: the eect on x` (p; w ) of a change in p`
or w .
I Engle/oer function: optimal bundles as a function of w or p` .
I Helpful to have Walrasian demand be continous and
dierentiable. Possible to establish that, when preferences are
continuous, strictly convex, locally nonsatiated on the
consumption set <L+ , the function x(p; w ) is continuous for all
(p; w ) 0.
Comparative statics: price eects

I The price oer curve x(p 1 ; : : : ; p` ; : : : ; p L ; w ) gives the

optimal bundles as a function of p` , for given prices p ` and
wealth w .
I The price eect of pk on the demand for good ` is given by
@x` (p;w )
the derivative @p k .
I Much more on this later...
I For the moment, satisfy yourself with the following
I when the own-price eect of good ` is positive, ` is a Gien
good (at the current price-wealth combination)
Comparative statics: price eects
Comparative statics: wealth eects 1
I For given prices p, the Engel function x(p; w ) gives the
optimal bundles as a function of wealth, for given prices.
I It can be represented by the wealth expansion path
Ep = fx(p; w ) : w > 0g (formally, its image in <L+ ).
Comparative statics: wealth eects 2

@x` (p;w )
The wealth eect for commodity ` is given by @w .
I commodity ` is normal at (p; w ) if its demand increases with
wealth
@x` (p;w )
@w 0 , NORMAL GOOD
I commodity ` is inferior at (p; w ) if its demand decreases with
wealth
@x` (p;w )
@w < 0 , INFERIOR GOOD
Wealth eect for special preferences

The wealth eect typically depends on the level of w .

However, for some types of preferences, this is not so:
I when preferences are homothetic, the income expansion path
is a straight line through the origin
,!the wealth eect is constant
I when preferences are quasi-linear, the wealth expansion path
is a straight horizontal line
,!the wealth eect is zero
UMP

indirect utility fct

From UMP, two interesting objects:
I optimal consumption bundles: the solution to UMP
,! consumers maximal utility value: the value function of the
UMP
The indirect utility function

The value function of the UMP is called the indirect utility

function v (p; w )
v (p; w ) = u(x ) = u(x(p; w ))
Properties of the indirect utility function:
1. homogeneous of degree zero in p and w
v (p; w ) = v ( p; w ) for any > 0
2. strictly increasing in w and non increasing in p`
3. quasi-convex in p
the lower contour set is convex
Price indierence curves
Inverting the indirect utility function
Since strictly increasing, can invert v (p; w ) to give the minimum
level of income necessary to reach a given level of utility.

Formally, this may be stated as the ...

The expenditure minimisation problem

EMP
The consumers decision problem may also be described as

min p x s.t.:u u EMP

x 0

In words,
- choose the least-cost consumption bundle
which ensures utility u
EMP is the dual problem of UMP
- it reverses the role of the objective function and of the constraint
The expenditure minimisation problem
Relationship between EMP e UMP

Strict relationship between EMP e UMP

I if x solves UMP for w
~ , then x solves EMP for u = u(x )
and p x = w ~
I if x solves EMP for u~, then x solves UMP when p x = w
and u = u(x )
Moving around objects
EMP

From EMP, two interesting objects:

I optimal consumption bundles: the solution to EMP
I consumers minimal expenditure: the value function of the
EMP
EMP

expenditure fct
From EMP, two interesting objects:
I optimal consumption bundles: the solution to EMP
,! consumers minimal expenditure: the value function of the
EMP
The expenditure function

The value function of the EMP is called the expenditure function

e(p; u).
If denote with x any solution to EMP, then
e(p; u) = p x
Properties of the expenditure function (analogous to those of the
indirect utility function)
1. homogeneous of degree 1 in p: e( p; u) = e(p; u) for any
>0
2. strictly increasing in u and non decreasing in p`
3. continous in p and u
4. concave in p
Concavity of the expenditure function

I Most important property of the expenditure function is the

concavity in prices
I Intuition: Let x solve EMP when prices are p
If p1 changes and x doesnt
I total expenditure varies linearly with p1
I total expenditure need to be higher or equal to total
expenditure with optimal behaviour (as from e(p; u))
Concavity of the expenditure function
Concavity of the expenditure function
Relationship between expenditure and indirect utility
functions

The relation between the solutions of the UMP and the EMP
implies that
e(p; v (p; w )) = w and v (p; e(p; u)) = u
Further consequence is that, for a given price vector
e(p; u) and v (p; w ) are inverses to one another
In other words, can solve
I w = e(p; u) for u to have the indirect utility fct v (p; w )
I u = v (p; w ) for w to have the expenditure fct e(p; u)
Moving around objects
EMP

hicksian demand
From EMP, two interesting objects:
,! optimal consumption bundles: the solution to EMP
I consumers minimal expenditure: the value function of the
EMP
Hicksian demand
Necessary conditions for solution to EMP

I The set of optimal commodity vector in EMP is h(p; u) and is

known as Hicksian (or compensated) demand correspondence
(or function)
I If h 2 h(p; u) is a solution to EMP, then there exists a
Lagrangean multiplier such that, for all `,
p ru(h ) h (p ru(h )) = 0
Why compensated??
Properties of Hicksian demand correspondence

Properties of Hicksian demand correspondence are

1. homogeneity of degree zero in p:
h( p; u) = h(p; u) for any > 0
2. no excess utility: for any x 2 h(p; u), then u(x) = u
3. h(p; u) is convex
I if is convex (i.e. u(x) is quasi-concave), then h(p; u) is a
convex set
I if is strictly convex (i.e. u(x) is strictly quasi-concave), then
h(p; u) is a single element
Compensated law of demand

For all p 0 and p 00 0,

00 0
(p p ) [h(p 00 ; u) h(p 0 ; u)] 0
) own-price eects are non positive

Always true for Hicksian demand but not necessarily the case for
Walrasian demand
Law of demand for Walrasian demand
Law of demand for Hicksian demand
Relationship between Walrasian and Hicksian demand

Clear relationship between Walrasian and Hicksian demand

correspondence

x(p; w ) = x(p; e(p; u)) = h(p; u)

and
h(p; u) = h(p; v (p; w )) = x(p; w )

These relationships have an operational content

,! how to obtain one type of demand from the other
Moving around objects
Relationships between objects

On the existing relationships between the dierent objects studied

so far
I Hicksian demand and expenditure function
I Hicksian demand and Walrasian demand
I Walrasian demand and indirect utility function
Relationships between objects

On the existing relationships between the dierent objects studied

so far
,! Hicksian demand and expenditure function
I Hicksian demand and Walrasian demand
I Walrasian demand and indirect utility function
Shepards lemma 1
Notice that:
rp e (:) = rp p h (p; u)
m 1 1 m m 1

by the denition of e (:)

T
= h (p; u) + p Dp h (p; u)
1 m m m

by the chain rule

= h (:) + [ru (h (:)) Dp h (p; u)]T

m 1 m 1

using p = ru (h (p; u)) from FOC-UMP

= h (p; u)

dierentiatig w.r.t. p constraint in EMP u (h (p; u)) = u

Shepards lemma 2

In words,

a change in prices has two eects on optimal total expenditure

I h(p; u) direct eect holding demand xed
I [p Dp h(p; u)] indirect eect due to the induced change in
demand holding prices xed

The second indirect eect cancels out since bundles always

minimise costs
Price derivatives of Hicksian demand

I The relationship between Hicksian demand and expenditure

function has consequences on the matrix of price derivatives
of Hicksian demand.
I Let Dp h(p; u) denote the L L matrix of rst price derivative
of Hicksian demands. Then
1. Dp h(p; u) = Dp2 e(p; u)
2. Dp h(p; u) is symmetric
3. Dp h(p; u) is negative semi-denite
A matrix is Negative Semi-denite if the determinants of all of its
principal submatrices are alternate in sign, starting with a negative
(with the allowance here of 0 determinants replacing one or more of
the positive or negative values)
I 1. and 2. are natural consequences of the Shepards lemma;
I 3. follows from the concavity of e(p; u) and implies the
compensated law of demand
Relationships between objects

On the existing relationships between the dierent objects studied

so far
I Hicksian demand and expenditure function
,! Hicksian and Walrasian demand
I Walrasian demand and indirect utility function
Hicksian and Walrasian demand

Slutsky equation
h(:) not observable, so?

Can obtain its derivatives from the observable x(p; w ) using the
)Slutsky equation(

@h` (p; u) @x` (p; w ) @x` (p; w )

= + xk (p; w )
@pk @pk @w
Easy to obtain, simply dierentiating w.r.to p` the equality

h` (p; u) = x` (p; e(p; u))

and using Shepards lemma.
Consequences of Slutksy equation 1

Some important consequences of Slutksy equation

I Hicksian demand steeper than Walrasian demand when
commodities are normal
I matrix Dp h(p; u) observable
0 1
S11 : : : S1L
B .. .. .. C
B . C
Dp h(p; u) = S(p; w ) = B . . C
@ SL1 : : : SLL A
Relationships between objects

On the existing relationships between the dierent objects studied

so far
I Hicksian demand and expenditure function
I Hicksian and Walrasian demand
,! Walrasian demand and indirect utility function
On an alternative way of looking at consumers choices
WARP (NO INDIFFERENCE CURVES)
I Results similar if we base the analysis not on the existence of
a preference relation but instead on a requirement of
rationality of actual consumers choices.
I To be rational actual choices (walrasian demand) must satisfy
the weak axiom of revealed preference (WARP):
I Let (p 0 ; w 0 ) and (p 00 ; w 00 ) be any two price-wealth situations,
and let x 0 = x(p 0 ; w 0 ) and x 00 = x(p 00 ; w 00 ) be the choices
made the consumer in these situations (each bundle for each
budget set)
I WARP: If p 0 x(p 00 ; w 00 ) w 0 and x(p 0 ; w 0 ) 6= x(p 00 ; q 00 ),
then p 00 x(p 0 ; w 0 ) > w 00
I If bundle x 00 was available at price-wealth combination (p 0 ; w 0 )
but was not chosen, it must be the case that x 0 is preferred to
x 00 . Then, if x 00 is chosen at price-wealth combination
(p 00 ; w 00 ), it must be the case that x 0 is not available.
I in other words: for two choices x(p 0 ; w 0 ) 6= x(p 00 ; q 00 ) (one for
each budget set), we cannot have both p 0 x(p 00 ; w 00 ) w 0
and p 00 x(p 0 ; w 0 ) w 00 .
Weak axiom of revealed preference
Weak axiom of revealed preference

A choice theory based on WARP

I uses less structural assumptions,
I leads (almost) to the same results,
I almost??: it does not imply symmetry of the Slutsky matrix.
I not assuming the existence of an utility fct, WARP does not
allow to evaluate of consumers level of well-being
Welfare evaluation

Welfare
I May want to evaluate eect on utility of changes in the
economic environment, typically in prices.
I Utility and indirect utility functions are useful instrument.
I However, measure (but not sign) of the eect
I depends on functional forms.
I not comparable across individuals
Money metric indirect utility fct

I Money metric utility fct gives in money terms the eect on

utility of changes in p
I Take any two utility levels u 0 ; u 1 and a reference price vector
p.
e(p; u 0 ) and e(p; u 1 ) are money measures of
utility u 0 and u 1 at prices p.
I Assume now that u 0 and u 1 are optimal utility levels for price
p 0 and p 1 , so that u 0 = v (p 0 ; w ) and u 1 = v (p 1 ; w ). Then
I e(p; v (p 0 ; w )) and e(p; v (p 1 ; w )) are money measures of
indirect utility (wealth required to reach v (p 0 ; w ))
I e(p; v (p 1 ; w )) e(p; v (p 0 ; w )) is a money measure of the
welfare eect of the price change
Money metric indirect utility fct
Equivalent and Compensating variations

Welfare ranking of p1 and p0 : the consumer is better o under p1

i EC and CV are positive
Equivalent variation and hicksian demand

The Equivalent Variation has an immediate intepretation in terms

of Hicksian demand

Suppose only price of good 1 is varying. Since

e p 0 ; u 0 = e p 1 ; u 1 = w and h1 (p; u) = @e(p;u)
@p 1 , we can write:

EV (p 0 ; p 1 ; w ) = e(p 0 ; u 1 ) w

= e(p 0 ; u 1 ) e(p 1 ; u 1 )

R p10
= p 11
p11 ; p21 ; : : : ; pL1 ; u 1 )d p~11
h1 (~

EV measured by the area between the prices and to the left of the
Hicksian demand curve at u 1 .
Equivalent variation and hicksian demand
Compensating variation and hicksian demand

The Compensating Variation too has an immediate intepretation in

terms of Hicksian demand

Suppose only price of good 1 is varying.

CV (p 0 ; p 1 ; w ) = w e(p 1 ; u 0 )

= e(p 0 ; u 0 ) e(p 1 ; u 0 )

R p10
= p 11
p11 ; p21 ; : : : ; pL1 ; u 0 )d p~11
h1 (~

CV measured by the area between the prices and to the left of the
Hicksian demand curve at u 0 .
Compensating variation and hicksian demand
Equivalent vs compensating variations

I Clearly, EV and CV give dierent measures of the welfare

change.
I For normal goods,
EV (p 0 ; p 1 ; w ) > CV (p 0 ; p 1 ; w )
(recall the relative slope of the demand functions??)
I Not surprisingly, the two measures are identical when there
are not wealth eects, i.e. preferences are quasi-linear. In this
case,
h1 (p 1 ; u 0 ) = x1 (p 1 ; w ) = h1 (p 1 ; u 1 )
I the common value of EV and CV is the area between p10 and
p11 and to the left of the market (Walrasian) demand for good
1.