Documente Academic
Documente Profesional
Documente Cultură
Klaus M. Schmidt
LMU Munich
1 / 55
Definition 3.1
A preference relation on X is rational if it is complete and transitive.
2 / 55
Notation:
y x means that yl xl for all l {1, . . . , L}.
y > x means that yl xl for all l {1, . . . , L} and yk > xk for at least one
k {1, . . . , L}.
y x means that yl > xl for all l {1, . . . , L}
q
PL
2
k y x k=
l=1 (yl xl )
x2
.......
x2
...
...
...
...
...
...
...
...
...
...
...
...
...
..
...................................................................................................
Strong Monotonicity
x1
3 / 55
.......
...
...
...
...
...
...
...
...
...
...
...
...
...
...
..................................................................................................
Local Non-Satiation
x1
4 / 55
Remarks:
If preferences are strongly monotone, then a sufficient condition for the
consumer to be better off is that he gets more of one good without getting
less of any other good.
If preferences are monotone, then a sufficient condition for the consumer
to be better off is that he gets more of each good.
If preferences are locally non-satiated, then for each consumption bundle
x there exists another consumption bundle y in an -neighborhood of x
that is strictly preferred by the consumer. However, it is not sufficient that
y >> x.
Strong monotonicity implies monotonicity: If the consumer is better off if
y x, he is surely better off if y >> x.
Monotonicity implies local non-satiation: If the consumer is better off if
y >> x then there also exists a y in the -neighborhood of x (namely to
the northeast of x) such that y x.
Local non-satiation implies that indifference curves cannot be thick.
5 / 55
6 / 55
x2
......
w
...
...
...
...
...
...
...
...
...
...
...
...
...
...
.................................................................................................
..
Strictly Convex Preferences
x1
......
...
...
...
...
...
...
...
...
...
...
...
...
...
...
.................................................................................................
..
x1
7 / 55
Remarks:
Reminder: A function f : A R, defined on the convex set A RN is
quasiconcave iff
f (x) t and f (x 0 ) t implies that f (x + (1 )x 0 ) t
for any t R, x, x 0 A and (0, 1), i.e. if its upper contour sets
{x A : f (x) t} are convex sets.
Reminder: A function f : A R, defined on the convex set A RN is
concave iff
f (x + (1 )x 0 ) f (x) + (1 )f (x 0 )
for any x, x 0 A and (0, 1).
Concavity implies quasi-concavity.
Quasi-concavity does not imply concavity: Quasi-concavity is an ordinal
property that survives any positive, monotone transformation, which is
not true for concavity.
8 / 55
Two Examples:
1. Homothetic Preferences:
A preference relation is homothetic iff x y implies x y for any
> 0. Note that for homothetic preferences the wealth expansion paths
are straight lines starting from the origin.
x2
.......
...
...
...
...
...
...
...
...
...
...
...
...
...
...
..................................................................................................
x1
9 / 55
2. Quasilinear Preferences
A preference relation is quasilinear iff x y implies
(x + e1 ) (y + e1 ) for e1 = (1, 0, . . . , 0) and any > 0.
Note that with quasilinear preferences the indifference sets are parallel
displacements of each other along the axis of commodity 1.
x2
......
...
...
...
...
...
...
...
...
...
...
...
...
...
...
..................................................................................................
.
x1
10 / 55
Definition 3.4
The preference relation is continuous iff for any sequence of pairs
n
n
n
n
{x n , y n }
n=0 with x y for all n, x = limn x and y = limn y , we have
x y.
Remarks:
An alternative and equivalent definition of continuity requires that the
upper and lower contour sets of must be closed (contain their
boundaries).
Continuity implies that indifference curves cannot have jumps. Show this
using the definition of continuity.
Consider the lexicographic preference relation in the two goods case, i.e.,
X = R2+ and x y if and only if x1 > y1 or x1 = y1 and x2 y2 . Are these
preferences continuous?
x2
.......
w
...
...
...
...
...
...
...
...
...
...
...
...
...
..
...................................................................................................
Continuity 1
x1
11 / 55
.......
...
...
...
...
...
...
...
...
...
...
...
...
...
...
..................................................................................................
Continuity 2
x1
12 / 55
Proposition 3.2
Suppose that X = RL+ . If is rational, monotone and continuous. Then there
is a utility function u(x) that represents .
Proof sketch: We first give a graphical sketch of the proof. The proof is
constructive, i.e., it shows how a utility function that represents can be
constructed if is rational, monotone and continuous.
x2
13 / 55
......
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...........................................................................................................................
x1
14 / 55
15 / 55
Remarks:
1. Recall that a utility function is unique only up to a positive, monotone
transformation.
2. A much stronger proposition holds but is more difficult to prove: We do
not have to require that X = RL+ nor that is monotone. Furthermore, it
can be shown that there is a continuous utility function that represents .
However, not all utility functions that represent have to be continuous.
Why not?
3. It is often convenient to work with a differentiable utility function.
However, there are continuous preferences that cannot be represented
by a differentiable utility function. Example?
Thus, in order to guarantee the existence of a differentiable utility function
we need an additional assumption on the smoothness of preferences.
16 / 55
s.t. p x w
Note that p = (p1 , . . . , pL ) is a row vector.
Proposition 3.3
If p 0 and u() is continuous, then the utility maximization problem has a
solution.
17 / 55
18 / 55
Demand Correspondence
Let x(p, w) be the solution to UMP. This is called the Walrasian (sometimes
also Marshallian) demand correspondence. If x(p, w) is unique for all
(p, w), then it is called the demand function.
Proposition 3.4
The demand correspondence x(p, w) satisfies the following properties:
(a) Homogeneity of degree 0 in (p, w)
(b) Walras Law
(c) If is strictly convex (so that u() is strictly quasiconcave), then x(p, w) is
unique for all (p, w).
It can also be shown that x(p, w) is continuous (upper hemicontinuous if
x(p, w) is a correspondence). Furthermore, if x(p, w) is a demand function,
then, under a mild additional assumption, it is also differentiable.
19 / 55
20 / 55
21 / 55
3. The UMP has a solution. The solution must satisfy Lagranges condition.
Hence, if there is only one (x , ) satisfying Lagranges condition, then it
must be the solution.
4. Interpretation of ?
5. Interpretation of
[u(x ) p] x = 0 ?
u
u
Notation: u(x ) = ( x
, . . . , x
).
1
L
22 / 55
Proposition 3.5
The indirect utility function v (p, w) satisfies the following properties:
(a) Homogeneity of degree 0 in (p, w)
(b) Strictly increasing in w
(c) Nonincreasing in pl for any l
23 / 55
24 / 55
l
xl (p, w) = v (p,w)
.
25 / 55
26 / 55
Remarks:
1. The proof is an envelope theorem argument, but we did not use the
envelope theorem explicitly. When you solve the problems in the problem
sets or the exam, you should also not simply refer to the envelope
theorem but derive the result.
2. To interprete the result, it is useful to write
v (p, w)
v (p, m)
= xl (p, w)
pl
w
Suppose that price pl is reduced marginally by pl . This gives additional
income pl xl (p, w) to the consumer. The marginal utility of a Dollar is
(p,w)
. Hence, the price reduction
the same for all goods and equal to vw
(p,w)
increases the consumers utility by pl xl (p, w) vw
.
3. Roys Identy is very useful. If we know the indirect utility function, then it
is much easier to derive the demand function from indirect utility than
from the direct utility function.
27 / 55
s.t. u(x) u
The Expenditure Minimization Problem (EMP) is the dual problem to
UMP: It reverses the roles of the objective function and the constraint.
However, the EMP also characterizes the efficient use of resources by the
consumer. In fact, EMP and UMP are basically equivalent:
28 / 55
x2
......
x2
...
...
...
...
...
...
...
...
...
...
...
...
...
...
.................................................................................................
..
UMP
x1
......
...
...
...
...
...
...
...
...
...
...
...
...
...
...
.................................................................................................
..
EMP
x1
29 / 55
Proposition 3.7
Suppose that p 0.
(a) If x is optimal in UMP when wealth is w > 0, then x is optimal in EMP
when the required utility level is u(x ).
(b) If x is optimal in EMP when the required utility level is u, then x is
optimal in UMP when wealth is w = p x .
30 / 55
31 / 55
Proposition 3.8
For any p 0 the Hicksian demand correspondence h(p, u) has the following
properties:
(a) Homogeneity of degree 0 in p.
(b) No excess utility: For any x h(p, u), u(x) = u.
(c) If is strictly convex (so that u(x) is strictly quasiconcave) then there is a
unique solution to EMP for all (p, u).
Proof of Proposition 3.8: Analogous to the proof of corresponding
proposition on the indirect utility function.
32 / 55
x2
.......
...
...
...
...
...
...
...
...
...
...
...
...
...
...
..................................................................................................
x1
33 / 55
Proposition 3.9
Suppose that h(p, u) is single valued for all p 0. Then the Hicksian demand
function satisfies the compensated law of demand, i.e., for all p0 and p00
(p00 p0 ) [h(p00 , u) h(p0 , u)] 0 .
Remarks:
The proposition says that price changes and compensated demand
changes go in opposite directions.
In particular, if only one price goes up, the compensated demand for this
good goes down.
Note that the proposition refers to compensated demand changes, i.e.,
the consumer is compensated for the price change such that he can still
afford his old level of utility.
34 / 55
35 / 55
Proposition 3.10
The expenditure function e(p, u) is
(a) homogenous of degree 1 in p,
(b) strictly increasing in u,
(c) nondecreasing in pl for any l,
(d) concave in p.
Intuition for these results? Explain (d) graphically.
36 / 55
pl0 xl0
= e(p0 , u)
pl00 xl00
= e(p00 , u)
l=1
L
X
l=1
L
X
pl x l
= e(p, u)
l=1
37 / 55
pl0 x l
l=1
L
X
L
X
pl0 xl0
l=1
pl00 x l
L
X
pl00 xl00
l=1
l=1
Multiplying the first inequality with and the second one with (1 ) and
adding up both inequalities yields:
L
X
[pl0 x l
l=1
+ (1
)pl00 x l ]
L
X
pl0 xl0
l=1
L
X
+
(1 )pl00 xl00
l=1
38 / 55
Hence:
e(p, u) =
L
X
pl x l
l=1
L
X
l=1
pl0 xl0 +
L
X
(1 )pl00 xl00
l=1
e(p0 , u) + (1 )e(p00 , u)
Q.E.D.
e(p, u)
39 / 55
......
...
....
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
.........................................................................................................................
p1
40 / 55
The expenditure function can easily be derived from the Hicksian demand
function by e(p, u) = p h(p, u). However, the following result shows that we
can also derive the Hicksian demand function from the expenditure function:
e(p, u)
pl
for all l = 1, . . . , L.
41 / 55
42 / 55
Remarks:
1. The proof is an envelope theorem argument, but we did not use the
envelope theorem explicitly.
2. Interpretation: In finite approximation Shephards Lemma says:
e(p, u) = hl (p, u) pl
If the price for good l increases by pl , then the consumer has to pay
pl hl (p, u) in addition, in order to buy the same consumption bundle that
he consumed before the price change. If pl is very small, then this is
also the amount necessary to get back to the old utility level u. The
reason is that for small price changes and starting from an optimally
chosen consumption bundle the substitution effects can be ignored.
3. Shephards Lemma is very useful. If we know the expenditure function, it
is much simpler to derive the Hicksian demand function via e(p, u) than
to derive it from the direct utility function.
43 / 55
Proposition 3.12
Suppose that h(p, u) is single valued and continuously differentiable at (p, u),
and denote its L L derivative matrix by Dp h(p, u). Then
(a) Dp h(p, u) = Dp2 e(p, u),
(b) Dp h(p, u) is a negative semidefinite matrix,
(c) Dp h(p, u) is a symmetric matrix,
(d) Dp h(p, u)p = 0.
44 / 55
45 / 55
Remarks:
1. Negative semidefiniteness of Dp h(p, u) is again the compensated law of
l (p,u)
demand. In particular, it implies that hp
0, i.e., the own substitution
l
effect is non-positive.
2. Symmetry of Dp h(p, u) requires that
hl (p, u)
hk (p, u)
=
pk
pl
This is not obvious and it is difficult to give an intuitive explanation for this
unexpected result.
l (p,u)
l (p,u)
3. If hp
0 then goods l and k are substitutes. If hp
< 0, then they
k
k
are called complements. Property (d) implies that for each good there
l (p,u)
0.
exists at least one substitute. This follows immediately from hp
l
46 / 55
47 / 55
48 / 55
Remarks:
1. The Slutsky Equation shows how the properties of the unobservable
Hicksian demand function translate to the observable Walrasian demand
function.
2. In particular, the Slutsky Equation implies
xl (p, w) xl (p, w)
hl (p, u)
=
+
xl (p, w) .
pl
pl
w
3. If good l is a normal good (xl /w > 0), then an increase in pl reduces
the Hicksian demand by less than the (Walrasian) demand. In the usual
demand diagram (with p on the vertical axis), the Hicksian demand
function is steeper than the (Walrasian) demand function.
4. If good l is an inferior good (xl /w < 0), then an increase in pl reduces
the Hicksian demand by more than the (Walrasian) demand. In the usual
demand diagram (with p on the vertical axis), the Hicksian demand
function is less steep than the (Walrasian) demand function. A good can
be a Giffen good only if it is inferior.
49 / 55
......
......
...
...
..
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
...
.
................................................................................................... ...................................................................................................
x
50 / 55
UMP
EMP
x(p, w)
h(p, u)
v (p, w)
e(p, u)
51 / 55
52 / 55
We started out from the assumptions that the consumers preferences are
rational, locally non-satiated and continuous, so they can be represented by a
utility function. We have shown that these assumptions have the following
implications for the (Walrasian) demand function x(p, w):
1. Homogeneity of degree 0
2. Walras Law
3. Compensated Law of Demand (Slutsky matrix is negative
semi-definite)
4. Symmetry of the Slutsky matrix
These restrictions on Walrasian demand function are rather weak and do not
impose much structure on observable demand.
53 / 55
A natural question is, whether there are any other properties of the demand
function that are implied by these assumptions.
The answer is no. It can be shown that for any demand function that satisfies
(1) to (4) there exists utility function (representing a rational preference
relation) such that the demand function is generated by this utility function.
This is known as the Integrability Problem (See MWH, Chapter 3.H). It also
shows, how the preferences of the consumer can (almost, but not quite) be
recovered from his observed demand behavior.
One could ask whether it is possible to impose an additional assumption in
the choice based approach that also yields a symmetric Slutsky matrix. This
assumption is the Strong Axiom of Revealed Preference.
54 / 55
Definition 3.5
The demand function x(p, w) satisfies the Strong Axiom of Revealed
Preference if for any list
(p1 , w 1 ), . . . , (pN , w N )
with x(pn+1 , w n+1 ) 6= x(pn , w n ) for all n N 1, we have pN x(p1 , w 1 ) > w N
whenever pn x(pn+1 , w n+1 ) w n for all n N 1.
This definition says, that if x(p1 , w 1 ) is directly or indirectly (through the chain
of x(pn , w n ) revealed preferred to x(pN , w N ), then x(pN , w N ) cannot be
revealed preferred to x(p1 , w 1 ), because x(p1 , w 1 ) is not affordable at
(pN , w N ).
With the Strong Axiom the choice based approach is equivalent to the
preference based approach.
55 / 55