Chapter 3: Classical Demand Theory: Frederic Vermeulen

Chapter 3: Classical demand theory
Frederic Vermeulen
University of Leuven
Department of Economics
2018-2019
F. Vermeulen (University of Leuven) Chapter 3 2018-2019 1 / 68

Introduction
Analysis of consumer demand by means of the preference-based

approach.
In Chapter 2, we focused on a choice-based approach: an individualís
choice behavior was the primitive characteristic of the individual.
Now, we will assume that a preference relation is the primitive
characteristic of the individual.
We will impose rationality axioms on these preferences and analyze
their implications on the choice behavior of an individual.
A number of objects that are of central interest in economics will be
introduced and the relationship between these will be analyzed. The
result is a coherent and beautiful theory that is also empirically very
relevant.

Overview
1 Preference relations: basic properties

2 Preference and utility
3 The utility maximization problem
4 The expenditure minimization problem
5 Relationships between demand, indirect utility and expenditure
functions
6 Integrability
7 Welfare evaluation of economic changes
8 The Strong Axiom of Revealed Preference

Overview

Preference relations: basic properties
The consumerís preferences over the commodity bundles in the

consumption set X ! RL+ are captured by a preference relation %.
This is a binary relation that allows the comparison of pairs of
alternatives x, y 2 X .
x % y : x is at least as good as y .
Strict preference relation #:
x # y () x % y and not y % x.
x # y : x is preferred to y .
Indi§erence relation &:
x & y () x % y and y % x.
x & y : x is indi§erent to y .

We will assume that the preference relation is rational.
DeÖnition
The preference relation % on X is rational if it possesses the following two
properties:
(i) Completeness: for all x, y 2 X , we have x % y or y % x.
(ii) Transitivity: for all x, y , z 2 X , if x % y and y % z, then x % z.
We will often assume that larger amounts of commodities are

preferred to smaller ones.
DeÖnition
The preference relation % on X is monotone if x 2 X and y ' x implies
y # x. It is strongly monotone if y ( x and y 6= x implies that y # x.

Often, the theory holds with the weaker assumption of local

nonsatiation.
DeÖnition
The preference relation % on X is locally nonsatiated if for every x 2 X
and every # > 0, there is some y 2 X such that kx + y k , # and y # x.
Three sets of consumption bundles can be derived:

The indi§erence set with respect to bundle x: fy 2 X jy & x g.
The upper contour set of bundle x: fy 2 X jy % x g.
The lower contour set of bundle x: fy 2 X jx % y g.

1 Local nonsatiation is implied by monotonicity of preferences. Because the converse isn't true, local
nonsatiation is a weaker condition.
2 There is no requirement that the preferred bundle y contain more of any good – hence, some goods
can be "bads" and preferences can be non-monotone.
3 It rules out the extreme case where all goods are "bads", since the point x = 0 would then be a bliss
point.
4 Local nonsatiation can only occur either if the consumption set is unbounded (open) (in other words, it
cannot be compact) or if x is on a section of a bounded consumption set sufficiently far away from the
ends. Near the ends of a bounded set, there would necessarily be a bliss point where local nonsatiation
does not hold.
Convexity is a central hypothesis in economics. It captures the idea of

diminishing marginal rates of substitution and the consumerís
inclination for diversiÖcation.
DeÖnition
The preference relation % on X is convex if for every x 2 X , the upper
contour set fy 2 X jy % x g is convex; that is, if y % x and z % x, then
ay + (1 + a) z % x for any a 2 [0, 1].
DeÖnition
The preference relation % on X is strictly convex if for every x 2 X , we
have that y % x, z % x, and y 6= z implies ay + (1 + a) z # x for all
a 2 (0, 1).

Overview


Preference and utility
The analysis of consumer demand is much easier if the preference

relation is represented by a utility function.
DeÖnition
A function u : X ! R is a utility function that represents the preference
relation % if 8x, y 2 X : x % y () u (x ) ( u (y ) .
The assumptions made thus far are not su¢cient to guarantee the
existence of a utility function that represents some rational preference
relation.

Lexicographic preferences or lexicographic orderings describe comparative preferences where an economic agent prefers any
amount of one good (X) to any amount of another (Y). Specifically, if offered several bundles of goods, the agent will choose the
bundle that offers the most X, no matter how much Y there is. Only when there is a tie between bundles with regard to the number of
units of X will the agent start comparing the number of units of Y across bundles.
A notorious example is a lexicographic preference relation: x % y if

either ìx1 > y1 î or ìx1 = y1 and x2 ( y2 î (for X = R2+ ).
The lexicographic preference relation is complete, transitive, strongly
monotone and strongly convex (see Exercise 3.C.1).
However, a utility function that represents these preferences does not
exist. No two di§erent bundles are indi§erent to each other: all
indi§erence sets are singletons. There are two dimensions of
indi§erence sets (one related to x1 > y1 and one related to x1 = y1
and x2 ( y2 ). Each of the indi§erence sets must be assigned an order
preserving utility number from the one-dimensional real line. This is
impossible since there are ënot enough real numbersí.
no two indifferent to each other: cause there always some preference statement true
if not x1 > y1, then maybe x1 < y1, but then y1 > x1 and so y1 preferred to x1 etc.

We assumed earlier that the preference relation % is rational. This is
a necessary condition for the representability by a utility function.
Theorem
A preference relation % can be represented by a utility function only if it is
rational.
Proof. We will show that if a utility function that represents % exists,
then % is complete and transitive.
(i) Completeness: Since u is a real-valued function on X , we have that for
any x, y 2 X either u (x ) ( u (y ) or u (y ) ( u (x ). Moreover, since u
represents %, this implies that either x % y or y % x. Hence, % is
complete.
(ii) Transitivity: Suppose x % y and y % z. Since u represents %, we have
u (x ) ( u (y ) and u (y ) ( u (z ) and, thus, u (x ) ( u (z ). Since u
represents %, we have x % z. So we have that x % y and y % z implies
x % z, which establishes transitivity."
The existence of a utility function is ensured if the preference relation

is continuous.
DeÖnition
The preference relation % is continuous if the following holds. For any
sequence of pairs f(x n , y n )gn•=1 with x n % y n for all Önite n, we have
x % y for x = limn !• x n and y = limn !• y n .
Continuity implies that the preferences cannot exhibit ëjumpsí.

Lexicographic ! preferences
" are not!continuous:
" Take the sequence of
1
0
bundles x n = n and y n = . For every n, we have
0 1
x!n % "
y n . The limits of both sequences
! " are equal to x = limn !• x =
n
0 0
and y = limn !• y n = . We thus obtain y # x, which
0 1
does not satisfy the deÖnition of continuity.
Theorem
Suppose that the (rational) preference relation % on X is continuous.
Then there is a continuous utility function u (x ) that represents %.
Proof. Interested students are referred to Mas-Colell et. al. (1995)."

Let us assume that the consumerís preference relation is continuous

implying that it can be represented by a continuous utility function.
This utility function is not unique: any strictly increasing
transformation of u (.), denoted by v (x ) = f (u (x )) for all x 2 X
and where f 0 (.) > 0, also represents %. So the utility value attached
to a particular indi§erence curve is a cardinal property (and not an
ordinal property)!
Example: The utility functions x1a x21 +a and a ln x1 + (1 + a) ln x2
represent the same preferences (a 2 (0, 1) and x ' 0)
(Cobb-Douglas preferences).

In what follows, we will often assume that u (.) is di§erentiable.

Monotonicity of the preference relation implies that u (.) is increasing
(which is an ordinal property).
The (strict) convexity of preferences implies that u (.) is (strictly)
quasiconcave (which is an ordinal property).
DeÖnition
A
# utilityL function is (strictly)
$ quasiconcave if the set
y 2 R + : u (y ) ( u (x ) is (strictly) convex for all x. Or, equivalently, if
u (ax + (1 + a) y ) ( (>) min fu (x ) , u (y )g for all x, y and a 2 [0, 1]
(x 6= y and a 2 (0, 1)).
https://mjo.osborne.economics.utoronto.ca/index.php/tutorial/index/1/qcc/t
quasiconcavity: every set of points inside the contour line of a function is convex

Overview


The utility maximization problem
In what follows, we assume that the consumer has a rational,

continuous and locally nonsatiated preference relation over the
bundles x 2 X = RL+ and assume that u (.) is a continuous utility
function that represents these preferences.
The consumerís decision problem is to choose the most preferred
consumption bundle given prices p ' 0 and wealth level w > 0
(throughout, we take these constraints as given).
This can be represented by the utility maximization problem (UMP):
max u (x )
x (0
subject to p 3 x , w .

Theorem
If u (.) is continuous, then the UMP has a solution.
# $
Proof. The Walrasian budget set Bp,w = x 2 RL+ : p 3 x , w is a
compact set because it is both bounded (for any `, we have x` , w /p` for
all x 2 Bp,w ) and closed. A continuous function always has a maximum
value on any compact set."
Two important objects are implied by the UMP: the set of optimal
commodity bundles (the maximizers / solution set of the UMP) and
the consumerís maximal utility value (the value function of the UMP).

The rule that assigns the set of optimal commodity bundles to each
price-wealth situation (p, w ) is x (p, w ) 2 RL+ and is called the
Marshallian (ordinary or Walrasian) demand correspondence. If
x (p, w ) is single-valued, we obtain the Marshallian (ordinary or
Walrasian) demand function.
Theorem
Suppose that the utility function u (.) is a continuous utility function
representing a locally nonsatiated preference relation % deÖned on the
consumption set X = RL+ . Then the Marshallian demand correspondence
x (p, w ) possesses the following properties:
(i) Homogeneity of degree zero in (p, w ): x (ap, aw ) = x (p, w ) for any
p, w and scalar a > 0.
(ii) Walrasí law (adding-up): p 3 x = w for all x 2 x (p, w ).
(iii) Convexity/uniqueness: If % is convex (u (.) is quasiconcave), then
x (p, w ) is a convex set. If % is strictly convex (u (.) is strictly
quasiconcave), then x (p, w ) consists of a single element.
# $ # $
Proof. (i) Note that x 2 RL+ : p 3 x , w = x 2 RL+ : ap 3 x , aw
for any a > 0. Since neither the utility function nor the budget set
changes, the set of optimal bundles must be the same for both situations:
x (ap, aw ) = x (p, w ).
(ii) If p 3 x < w for some x 2 x (p, w ), then local nonsatiation implies that
there must be another bundle y su¢ciently close to x with both p 3 y < w
and y # x. But this would contradict the optimality of x.
(iii) Suppose that u (.) is quasiconcave and that there are two bundles x
and x 0 , with x 6= x 0 and x, x 0 2 x (p, w ). We will now show that
x 00 = ax + (1 + a) x 0 2 x (p, w ) for any a 2 [0, 1]. Note that
u (x ) = u (x 0 ), which is denoted by u 4 . By quasiconcavity, u (x 00 ) ( u 4 .
Since p 3 x , w and p 3 x 0 , w , we have
p 3 x 00 = p 3 [ax + (1 + a) x 0 ] , w . So x 00 is a feasible choice. Since
u (x 00 ) ( u 4 and x 00 is feasible, we have x 00 2 x (p, w ), which establishes
convexity of x (p, w ) when u (.) is quasiconcave.

cause it doesnt maximize u
Suppose that u (.) is strictly quasiconcave, then with the same reasoning
as before, we obtain that x 00 is a feasible choice and that u (x 00 ) > u (x 4 )
for a 2 (0, 1). This contradicts that x, x 0 2 x (p, w ), which establishes
uniqueness when u (.) is strictly quasiconcave."

If u (.) is continuously di§erentiable, then we can represent an

optimal consumption bundle x 4 2 x (p, w ) by means of the
Kuhn-Tucker (necessary) conditions associated with the UMP: if
x 4 2 x (p, w ) is a solution to the UMP, then there exists a Lagrange
multiplier l ( 0 such that for all ` = 1, ..., L:
∂u (x 4 )
, lp` , with equality if x`4 > 0.
∂x`
Or,
5u (x 4 ) , lp
and
x 4 3 [5u (x 4 ) + lp ] = 0.

If there is an interior optimum (x 4 ' 0) and 5u (x 4 ) ' 0, then the

Kuhn-Tucker conditions imply that for any two commodities ` and k,
we have
∂u (x 4 )
∂x ` p
∂u (x 4 )
= `.
pk
∂xk
This shows that the marginal rate of substitution between any two
goods is equal to their price ratio at an interior optimum.

The nonnegative Lagrange multiplier l gives the marginal (shadow)

value of relaxing the budget constraint in the UMP; it equals the
consumerís marginal utility of wealth.
Suppose that x (p, w ) is di§erentiable and strictly positive. The
change in utility following a marginal increase in w equals
5u (x (p, w )) 3 Dw x (p, w ) = lp 3 Dw x (p, w ) = l.
The necessary Kuhn-Tucker conditions are also su¢cient if the utility
function u (.) is quasiconcave and monotone and that 5u (x ) 6= 0 for
all x 2 RL+ .

For each (p, w ) ' 0, the utility value of the UMP is denoted
v (p, w ) 2 R and equals u (x 4 ) for any x 4 2 x (p, w ).
The function v (p, w ) is the indirect utility function.
The indirect utility function depends on the utility representation
chosen.
Theorem
Suppose that u (.) is a continuous function that represents a locally
nonsatiated preference relation % on the consumption set X = RL+ . The
indirect utility function v (p, w ) then satisÖes the properties:
(i) Homogeneous of degree zero.
(ii) Nondecreasing in w and nonincreasing in p` for any `.
(iii) Quasiconvex; i.e., the set f(p, w ) : v (p, w ) , v g is convex for any v .
(iv) Continuous in p and w .

The functions and their variables are different, so there is no "inflection" or flipping.
The utility function u which maps from the space of goods X to ℝ is convex and quasiconcave.
The indirect utility function v which maps from the space of prices to ℝ is quasiconvex.
Intuitively:
u: If you average two consumption bundles your utility is not lower than the average of the utility of the two bundles. Rather than
eating just meat one day and just vegetables the other day you prefer to mix these everyday.
v: If the price vector is p one day and p′ the other day you may be better off than if it was p+p′2 every day. It is easy to check that
anything you can buy under the second price regime you can also buy under the first. However there might be consumption bundles
that you can only buy under the first price regime.
Proof. Interested students are referred to Mas-Colell, Whinston and Green

(1995)."

Overview


The expenditure minimization problem
Throughout, we assume that u (.) is a continuous utility function that

represents locally nonsatiated preferences over the consumption set
X = RL+ . We further assume that u > u (0) and p ' 0.
The expenditure minimization problem (EMP) is deÖned as:
min p 3 x
x (0
subject to u (x ) ( u.
It looks for the minimum level of wealth needed to reach a utility level
that is at least equal to u.
The EMP is known as the ëdualí problem of the UMP: the objective
function and the budget constraint play reversed roles in both
optimization problems.

Theorem
Suppose that u (.) is a continuous utility function that represents a locally
nonsatiated preference relation % deÖned on the consumption set X = RL+
and that the price vector is p ' 0. Then we have
(i) If x 4 is optimal in the UMP when wealth w > 0, then x 4 is optimal in
the EMP when the required utility level is u (x 4 ). The minimized
expenditure level in this EMP is exactly w .
(ii) If x 4 is optimal in the EMP when the required utility level is u > u (0),
then x 4 is optimal in the UMP when wealth is p 3 x 4 . The maximized
utility level in the UMP is exactly u.

Proof. (i) Suppose that x 4 is not optimal in the EMP with required utility
level u (x 4 ). Then there exists an x 0 such that u (x 0 ) ( u (x 4 ) and
p 3 x 0 < p 3 x 4 , w . By local nonsatiation, there is an x 00 close to x 0 such
that u (x 00 ) > u (x 0 ) and p 3 x 00 < w . This implies that x 00 2 Bp,w and
u (x 00 ) > u (x 4 ), which contradicts the optimality of x 4 in the UMP. So x 4
is also optimal in the EMP with the required utility level u (x 4 ). The
minimal wealth level equals p 3 x 4 . Since x 4 satisÖes Walrasí law, we have
p 3 x4 = w.
(ii) Since u > u (0), we have x 4 6= 0. Hence, p 3 x 4 > 0. Suppose that x 4
is not optimal in the UMP when wealth equals p 3 x 4 . Then there exists an
x 0 such that u (x 0 ) > u (x 4 ) and p 3 x 0 , p 3 x 4 . Consider a bundle
x 00 = ax 0 where a 2 (0, 1). By continuity of u (.), we will have
u (x 00 ) > u (x 4 ) and p 3 x 00 < p 3 x 4 if a is close enough to 1. This however
contradicts the optimality of x 4 in the EMP. As a result, x 4 is optimal in
the UMP when w = p 3 x 4 while the maximal utility level equals u (x 4 ).
Further, we will show that u (x 4 ) = u."
The value of the EMP is denoted e (p, u ) and is known as the

expenditure (cost) function.
The value of the expenditure function for any (p, u ) is equal to p 3 x 4
for any optimal x 4 .
Theorem
Suppose that u (.) is continuous and representing a locally nonsatiated
preference relation % deÖned on the consumption set X = RL+ . The
expenditure function e (p, u ) is
(i) Homogeneous of degree one in p.
(ii) Strictly increasing in u and nondecreasing in p` for any `.
(iii) Concave in p.

if all prices change proportionally, to reach the same utility level u we will still want to
reach the same bundle x, but now the wealth we need for that will be larger
Proof. (i) The constraint set of the EMP is unchanged when the prices
change. So, for any scalar a > 0, minimizing (ap ) 3 x on this set obtains
the same optimal consumption bundles as minimizing p 3 x. Denote an
optimal bundle by x 4 . We thus have e (ap, u ) = ap 3 x 4 = ae (p, u ).
(ii) Suppose that e (p, u ) is not strictly increasing in u and let x 0 and x 00
denote optimal consumption bundles for required utility levels u 0 and u 00
where u 00 > u 0 and p 3 x 0 ( p 3 x 00 > 0. Consider e x = ax 00 , where
a 2 (0, 1). By continuity of u (.), there exists an a close enough to 1 such
that u (e x ) > u 0 and p 3 x 0 > p 3 e x . This contradicts x 0 being optimal in the
EMP with required utility level u 0 . Next, suppose that the price vectors p 00
and p 0 have p`00 ( p`0 and pk00 = pk0 for all k 6= `. Let x 00 be an optimizing
bundle in the EMP for prices p 00 . Then
e (p 00 , u ) = p 00 3 x 00 ( p 0 3 x 00 ( e (p 0 , u ).

Proof. (iii) Fix a required utility level u and let p 00 = ap + (1 + a) p 0 for

a 2 [0, 1]. Suppose that x 00 is optimal in the EMP when prices are p 00 . We
then have
& '
e p 00 , u = p 00 3 x 00
= ap 3 x 00 + (1 + a) p 0 3 x 00
& '
( ae (p, u ) + (1 + a) e p 0 , u
because u (x 00 ) ( u, p 3 x 00 ( e (p, u ) and p 0 3 x 00 ( e (p 0 , u )."

its bigger than or equal to simply because its not the optimal bundle, and the
optimal bundle minimizes

From above, it should be clear that there is a close relation between

the expenditure function and the indirect utility function for any
p ' 0, w > 0 and u > u (0):
e (p, v (p, w )) = w
v (p, e (p, u )) = u.
The relation implies that for a Öxed price vector, the expenditure
function and the indirect utility function are each otherís inverse (see
Exercises 3.E.8 and 3.E.9).

The set of optimal commodity bundles in the EMP is denoted by

h (p, u ) ! RL+ and is known as the Hicksian (compensated) demand
correspondence or function if h (p, u ) is single-valued.
Theorem
preference relation % deÖned on the consumption set X = RL+ . Then for
any p ' 0, the Hicksian demand correspondence h (p, u ) has the
properties
(i) Homogeneous of degree zero: h (ap, u ) = h (p, u ) for any a > 0 and
any (p, u ).
(ii) No excess utility: for any x 2 h (p, u ), u (x ) = u.
(iii) Convexity/uniqueness: if % is convex, then h (p, u ) is a convex set; if
% is strictly convex, then h (p, u ) is single-valued.

Proof. Interested students are referred to Mas-Colell, Whinston and Green

(1995)."

Also the Hicksian demand correspondences and the Marshallian

demand correspondences can be related to each other:
h (p, u ) = x (p, e (p, u ))

x (p, w ) = h (p, v (p, w )) .
This relation involves the Hicksian wealth compensation: the Hicksian

demand bundle shows how quantities change when prices change
while holding the utility level constant.
The Hicksian wealth compensation is the amount
4wHicks = e (p 0 , u ) + w .

In Chapter 2, we have seen that demand and prices move in opposite

directions for price changes that are accompanied by a Slutsky wealth
compensation. This is also the case for the Hicksian demand function.
An important implication of the following theorem is that for
compensated demand functions, the own-price e§ects are nonpositive
(recall that this was not true for Marshallian demand).
Theorem
preference relation % deÖned on the consumption set X = RL+ and that
h (p, u ) is single-valued for any p ' 0. Then the Hicksian demand
function h (p, u ) satisÖes the compensated law of demand: for all p 0 and
p 00 , & 00 ' ( & ' & ')
p + p 0 3 h p 00 , u + h p 0 , u , 0.

Proof. For any p ' 0, the bundle h (p, u ) is optimal in the EMP and so
it is associated with a lower expenditure level at prices p than any other
bundle that obtains a utility level of at least u. Therefore, we have
& ' & '
p 00 3 h p 00 , u , p 00 3 h p 0 , u
and & ' & '

p 0 3 h p 00 , u ( p 0 3 h p 0 , u .
Subtracting these equations yields the result."

Overview

functions

Relationships between demand, indirect utility and
expenditure functions
We again assume that u (.) is a continuous utility function that
represents a locally nonsatiated preference relation % deÖned on the
consumption set X = RL+ and we restrict attention to p ' 0.
We further assume a strictly convex preference relation so that
Marshallian and Hicksian demands are single-valued.
Theorem
(Shephardís lemma) Suppose that u (.) is a continuous utility function
that represents a locally nonsatiated and strictly convex preference relation
% deÖned on the consumption set X = RL+ . For all p and u, we have:
2 3
∂e (p,u )
∂p 1
6 .. 7
h (p, u ) = rp e (p, u ) = 6
4 . 7.
5
∂e (p,u )
∂p L

Proof. Assume that h (p, u ) ' 0 and that h (p, u ) is di§erentiable at

(p, u ). The change in expenditure is equal to
rp e (p, u ) = rp [p 3 h (p, u )]
= h (p, u ) + [p 3 Dp h (p, u )]T .
Substituting from the Örst-order conditions for an interior optimum to the

EMP, p = lOu (h (p, u )), yields
rp e (p, u ) = h (p, u ) + l [Ou (h (p, u )) 3 Dp h (p, u )]T .
Since u (h (p, u )) = u for all p, the term between brackets is equal to

zero, which proves the result."

Theorem
Suppose that u (.) is a continuous utility function that represents a locally
nonsatiated and strictly convex preference relation % deÖned on the
consumption set X = RL+ . Suppose also that h (., u ) is continuously
di§erentiable at (p, u ) and denote its L 8 L derivative matrix by
Dp h (p, u ). Then
(i) Dp h (p, u ) = Dp2 e (p, u ) .
(ii) Dp h (p, u ) is a negative semideÖnite and symmetric matrix.
(iii) Dp h (p, u ) p = 0.

Proof. (i) Di§erentiation of the equation h (p, u ) = Op e (p, u ) obtains

the result.
(ii) Property (i) implies that Dp h (p, u ) is the derivative matrix associated
with a twice continuously di§erentiable and concave expenditure function
e (p, u ). As a result, Dp h (p, u ) is symmetric and negative semideÖnite.
(iii) The demand function h (p, u ) is homogeneous of degree zero in p,
which implies that h (ap, u ) + h (p, u ) = 0 for any a > 0. Taking the
derivative of this expression with respect to a and evaluating at a = 1
obtains the result."

The negative-semideÖniteness of Dp h (p, u ) is the di§erential analog

of the compensated law of demand.
More speciÖcally, recall that negative-semideÖniteness implies that
dp 3 Dp h (p, u ) dp , 0 for any vector dp (see Chapter 2). Since
dh (p, u ) = Dp h (p, u ) dp, substitution obtains dp 3 dh (p, u ) , 0.
One of the implications of negative-semideÖniteness is that
∂h ` (p,u )
∂p ` , 0 for ` = 1, ..., L (cf. supra).
∂h (p,u ) ∂h (p,u )
Symmetry implies that for all k, ` we have `∂pk = k∂p` ; this is a
conclusion that we could not draw for the Marshallian demand.
∂h ` (p,u )
Two goods k and ` are substitutes (complements) if ∂p k ( (<) 0.
∂h ` (p,u )
Since , 0 and Dp h (p, u ) p = 0, each good ` = 1, ..., L has at
∂p `
least one substitute.

Theorem
(Slutsky equation) Suppose that u (.) is a continuous utility function that
represents a locally nonsatiated and strictly convex preference relation %
deÖned on the consumption set X = RL+ . Then for all (p, w ) and
u = v (p, w ), we have
∂h` (p, u ) ∂x (p, w ) ∂x` (p, w )

= ` + xk (p, w ) for all `, k
∂pk ∂pk ∂w
or
Dp h (p, u ) = Dp x (p, w ) + Dw x (p, w ) x (p, w )T .

Proof. Suppose that a consumer faces the price-wealth pair (p, w ) and
attains a utility level u. Note that w = e (p, u ). Recall that
h` (p, u ) = x` (p, e (p, u )) for all (p, u ). Di§erentiation with respect to pk ,
using Shephardís lemma and evaluation at (p, u ) obtains
∂h` (p, u ) ∂x (p, e (p, u )) ∂x` (p, e (p, u ))

= ` + hk (p, u ) .
∂pk ∂pk ∂w
Since w = e (p, u ) and hk (p, u ) = xk (p, w ), we obtain the result."

The above result implies that Dp h (p, u ) is equal to the Slutsky

matrix S (p, w ) that we derived in Chapter 2 (which, recall, is based
on a choice-based approach).
Since Dp h (p, u ) = S (p, w ), demand that is generated from
preference maximization must be associated with a Slutsky matrix
S (p, w ) that is negative semideÖnite, symmetric and satisfying
S (p, w ) p = 0.
In Chapter 2, we have shown that if x (p, w ) satisÖes WARP (and
homogeneity of degree zero and Walrasí law), then S (p, w ) is
negative semideÖnite with S (p, w ) p = 0.
This implies that the preference-based approach obtains stronger
restrictions than the choice-based approach built on WARP.

An extremely important implication from an empirical point of view is

that although the Hicksian demand function is not directly
observable, we can compute the Slutsky matrix by means of the
observable Marshallian demand function through the Slutsky equation
(see, e.g., Barten, 1969, Deaton and Muellbauer, 1980, and Banks,
Blundell and Lewbel, 1997, for some applications).
Once the Slutsky matrix S (p, w ) is known, the theoretical restrictions
can be tested to check the validity of the theory. Further, the Slutsky
matrix serves as an input in applications of optimal taxation theory
(see, e.g., Ramsey, 1927, Diamond and Mirrlees, 1971, Mirrlees,
1971).

Theorem
(Royís identity) Suppose that u (.) is a continuous utility function that
represents a locally nonsatiated and strictly convex preference relation %
deÖned on the consumption set X = RL+ . Suppose also that the indirect
utility function is di§erentiable at (p, w ) ' 0.Then we have for every
` = 1, ..., L
∂v (p,w )
∂p `
x` (p, w ) = + ∂v (p,w )
.
∂w

Proof. Let u = v (p, w ). Because the identity u = v (p, e (p, u )) holds

for all p, di§erentiation with respect to p` and evaluating the result at
p = p yields
∂v (p, e (p, u )) ∂v (p, e (p, u )) ∂e (p, u )

+ = 0.
∂p` ∂w ∂p`
∂e (p,u )
Shephardís lemma implies that ∂p` = h` (p, u ). We further know that
e (p, u ) = w , which implies that h` (p, u ) = x` (p, w ). Making the
appropriate substitutions and rearranging yields the result."

Royís identity is very important from an empirical point of view: it is

much easier to compute the Marshallian demand function from an
indirect utility function than from a direct utility function. In the
former case, one only needs derivatives. The latter case requires
solving a system of Örst-order conditions which can be cumbersome.
Well-known examples of demand systems that are based on an
indirect utility function are Deaton and Muellbauerís (1980) Almost
Ideal Demand System and Banks, Blundell and Lewbelís (1997)
Quadratic Almost Ideal Demand System.

Overview

functions
6 Integrability

Integrability
We already know that if a continuously di§erentiable demand

function x (p, w ) is generated by rational preferences, then the
demand function must satisfy Walrasí law, homogeneity of degree
zero and Slutsky symmetry and negativity.
We now analyze the reverse question ìIf we observe a demand
function x (p, w ) that has the above properties, can we Önd
preferences that rationalize the demand function?î
This is known as the integrability problem.
It turns out that the above conditions are indeed su¢cient to
guarantee the existence of a rational preference relation: if a demand
function satisÖes Walrasí law, homogeneity of degree zero and Slutsky
symmetry and negativity, then there exists a rational preference
ordering that generates demand.
Please do realize the beauty of the theory!

Integrability
This integrability result is very important:

The properties Walrasí law, homogeneity and Slutsky symmetry and
negativity are the only consequences of the preference-based approach.
As soon as demand satisÖes these properties, then there is some
rational preference relation that can rationalize demand.
Symmetry is the only property added to demand by the
preference-based approach beyond the implications of WARP, Walrasí
law and homogeneity.
To conduct welfare analyses (cf. infra), we need to know the
consumerís preferences. The integrability result implies that we can
recover these preferences on the basis of observed demand behavior.
An empirically fruitful approach is to start from a tractable demand
function and then check whether the above restrictions for a rational
preference relation are satisÖed.

Overview

functions
6 Integrability

Welfare evaluation of economic changes
Up to now, we analyzed the preference-based approach to consumer

behavior from a positive point of view.
Now, we will use the theory to analyze the e§ects of changes in the
consumerís environment on her well-being (which concerns the
normative side of the theory and which can be labelled as welfare
analysis).
This is very important from a policy point of view:
A policy maker may be interested in the impact on the consumerís
well-being of an introduction of a sugar tax, a change in the indirect
tax on gasoline (which impacts the consumer price) or of a change in
the wage tax (where leisure is a commodity).
The theory can also be used to construct index numbers like the true
cost-of-living index (see, e.g., Diewert, 1980, for a survey).
The theory is also used in the theory of optimal taxation (see, e.g.,
Ramsey, 1927, Diamond and Mirrlees, 1971, Mirrlees, 1971).

Suppose that we want to evaluate the impact on the consumerís

well-being of a change from p 0 (old prices) to p 1 (new prices).
Any indirect utility function that represents a consumerís preference
relation % allows to check whether a consumer is better or worse o§
after
& 1 the'price&change:
' the consumer is worse o§ if and only if
0
v p , w + v p , w < 0.
A particulary useful indirect utility function is Samuelsonís (1974)
money metric indirect utility function: it allows to express the change
in well-being in monetary units.
Start from any indirect utility function v (., .), take an arbitrary price
vector p ' 0 and consider the function e (p, v (p, w )), which gives
the minimal expenditures needed to reach the utility level v (p, w )
when the prices are equal to p.

Viewed as a function of (p, w ), e (p, v (p, w )) is nothing else than

another indirect utility function that represents the preference relation
% (note that e (p, v (., .)) is a strictly increasing transformation of
v (., .)).
& & '' & & ''
The di§erence e p, v p 1 , w + e p, v p 0 , w gives a measure of
the change in well-being expressed in monetary units.
Two natural choices for p are p 0 and p 1 , which lead to respectively
the equivalent variation (EV) and compensating variation (CV) that
were proposed by Hicks (1939).
0
& ' & '
& u0 =
Let ' v p 0&, w and' u 1 = v p 1 , w . Note that
e p , u0 = e p1 , u1 = w .

DeÖnition
Equivalent variation.
& ' & ' & '
EV p 0 , p 1 , w = e p0 , u1 + e p0 , u0
& '
= e p0 , u1 + w .
The EV is the monetary amount that makes the consumer just

indi§erent between accepting the amount (which is negative when
u 1 < u 0 ) instead of being faced with the price change from p 0 to p 1
(in both cases, the consumer would be on the new utility level u 1 ).

DeÖnition
Compensating variation.
& ' & ' & '
CV p 0 , p 1 , w = e p1 , u1 + e p1 , u0
& '
= w + e p1 , u0 .
The CV is the negative of the monetary amount that would just

compensate a consumer who is faced with the price change from p 0
to p 1 (this amount would bring the consumer back to the initial
utility level u 0 ).

Both measures have an interpretation in terms of the Hicksian
demand function.
Suppose that only the price of the Örst commodity changes such that
p10 6= p11 and p`0 = p`1 = p ` for all ` 6= 1.
& ' & ' ∂e (p,u )
Notice that w = e p 0 , u 0 = e p 1 , u 1 and h1 (p, u ) = ∂p1 and
denote p +1 = (p 2 , ..., p L ).
Consequently:
0 1 0 1 0 1
EV p 0 , p 1 , w = e p0 , u1 + e p1 , u1
Z p0 0 1
1
= h1 p1 , p +1 , u 1 dp1
p 11
0 1 0 1 0 1
CV p 0 , p 1 , w = e p0 , u0 + e p1 , u0
Z p0 0 1
1
h1 p1 , p +1 , u 0 dp1 .
p 11

The consumer is better o§ in the new situation if and only if the

measures EV and CV are positive. The speciÖc values will di§er in
general.
It can be shown that when there is no wealth e§ect for good 1 (i.e.,
∂x (p,w )
when 1∂w = 0 for all p, w ), then the values of EV and CV
associated with the change from p 0 to p 1 will be equal to the change
in the Marshallian consumer surplus that equals
R p10
p1 1
x (p1 , p +1 , w ) dp1 (Marshall, 1920).
1

Overview

functions
6 Integrability
8 The Strong Axiom of Revealed Preference

The Strong Axiom of Revealed Preference
It can be shown that consumer choice that satisÖes WARP cannot be
rationalized by a rational preference relation in general.
Houthakker (1950) showed that a strengthening of WARP provides a
necessary and su¢cient condition on consumer demand to be
rationalizable by a rational preference relation. This strengthening is
the so-called Strong Axiom of Revealed Preference (SARP).
DeÖnition
The Marshallian demand & function ' &x (p, w ) 'satisÖes SARP if for any series
of&price-wealth' pairs p 1 , w 1 , ..., p N , w N with
x p n +&1 , w n +1' 6= x (p n , w n ) for all n ,
& N + 1, we'have
p N 3 x p 1 , w 1 > w N whenever p n 3 x p n +1 , w n +1 , w n for all
n , N + 1.
& 1 1'
SARP tells that & Nif x Np' , w is&directly 'or indirectly revealed at least
as good as x &p , w ' , then x p , w N N cannot be directly revealed
preferred to x p 1 , w 1 .
Theorem
The Marshallian demand function x (p, w ) satisÖes SARP if and only if
there is a rational preference relation % that rationalizes x (p, w ) in the
sense that for all (p, w ), x (p, w ) # y for every y 6= x (p, w ) with
y 2 Bp,w .
Proof. Interested students are referred to Mas-Colell et. al. (1995)."

The theorem implies the remarkable conclusion that a choice-based

theory of demand that is founded on SARP is equivalent to the
preference-based theory of demand.
This implies that SARP is equivalent to the symmetry and negativity
of the Slutsky matrix.
It can be shown that WARP is equivalent to SARP (and thus Slutsky
symmetry) when there are only two commodities (L = 2) (see
Exercise 3.J.1). Hence, when L = 2 and demand satisÖes WARP,
then it can always be rationalized by a rational preference relation.

Chapter 3: Classical Demand Theory: Frederic Vermeulen

Uploaded by

Copyright:

Available Formats

Chapter 3: Classical Demand Theory: Frederic Vermeulen

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Chapter 3: Classical Demand Theory: Frederic Vermeulen

Uploaded by

Copyright:

Available Formats

Chapter 3: Classical demand theory

F. Vermeulen (University of Leuven) Chapter 3 2018-2019 1 / 68

Analysis of consumer demand by means of the preference-based

F. Vermeulen (University of Leuven) Chapter 3 2018-2019 2 / 68

1 Preference relations: basic properties

F. Vermeulen (University of Leuven) Chapter 3 2018-2019 3 / 68

1 Preference relations: basic properties

F. Vermeulen (University of Leuven) Chapter 3 2018-2019 4 / 68

The consumerís preferences over the commodity bundles in the

F. Vermeulen (University of Leuven) Chapter 3 2018-2019 5 / 68

We will assume that the preference relation is rational.

We will often assume that larger amounts of commodities are

F. Vermeulen (University of Leuven) Chapter 3 2018-2019 6 / 68

Often, the theory holds with the weaker assumption of local

Three sets of consumption bundles can be derived:

F. Vermeulen (University of Leuven) Chapter 3 2018-2019 7 / 68

Convexity is a central hypothesis in economics. It captures the idea of

F. Vermeulen (University of Leuven) Chapter 3 2018-2019 8 / 68

1 Preference relations: basic properties

F. Vermeulen (University of Leuven) Chapter 3 2018-2019 9 / 68

The analysis of consumer demand is much easier if the preference

F. Vermeulen (University of Leuven) Chapter 3 2018-2019 10 / 68

A notorious example is a lexicographic preference relation: x % y if

F. Vermeulen (University of Leuven) Chapter 3 2018-2019 11 / 68

The existence of a utility function is ensured if the preference relation

Continuity implies that the preferences cannot exhibit ëjumpsí.

Proof. Interested students are referred to Mas-Colell et. al. (1995)."

F. Vermeulen (University of Leuven) Chapter 3 2018-2019 14 / 68

Let us assume that the consumerís preference relation is continuous

F. Vermeulen (University of Leuven) Chapter 3 2018-2019 15 / 68

In what follows, we will often assume that u (.) is di§erentiable.

F. Vermeulen (University of Leuven) Chapter 3 2018-2019 16 / 68

1 Preference relations: basic properties

F. Vermeulen (University of Leuven) Chapter 3 2018-2019 17 / 68

In what follows, we assume that the consumer has a rational,

F. Vermeulen (University of Leuven) Chapter 3 2018-2019 18 / 68

F. Vermeulen (University of Leuven) Chapter 3 2018-2019 19 / 68

F. Vermeulen (University of Leuven) Chapter 3 2018-2019 21 / 68

cause it doesnt maximize u

F. Vermeulen (University of Leuven) Chapter 3 2018-2019 22 / 68

If u (.) is continuously di§erentiable, then we can represent an

F. Vermeulen (University of Leuven) Chapter 3 2018-2019 23 / 68

If there is an interior optimum (x 4 ' 0) and 5u (x 4 ) ' 0, then the

F. Vermeulen (University of Leuven) Chapter 3 2018-2019 24 / 68

The nonnegative Lagrange multiplier l gives the marginal (shadow)

F. Vermeulen (University of Leuven) Chapter 3 2018-2019 25 / 68

F. Vermeulen (University of Leuven) Chapter 3 2018-2019 26 / 68

Proof. Interested students are referred to Mas-Colell, Whinston and Green

F. Vermeulen (University of Leuven) Chapter 3 2018-2019 27 / 68

1 Preference relations: basic properties

F. Vermeulen (University of Leuven) Chapter 3 2018-2019 28 / 68

Throughout, we assume that u (.) is a continuous utility function that

F. Vermeulen (University of Leuven) Chapter 3 2018-2019 29 / 68

F. Vermeulen (University of Leuven) Chapter 3 2018-2019 30 / 68

The value of the EMP is denoted e (p, u ) and is known as the

F. Vermeulen (University of Leuven) Chapter 3 2018-2019 32 / 68

F. Vermeulen (University of Leuven) Chapter 3 2018-2019 33 / 68

Proof. (iii) Fix a required utility level u and let p 00 = ap + (1 + a) p 0 for