Week 9: Topics in Consumer Theory (Jehle and

Reny, Chapter 2)
Tsun-Feng Chiang
*School of Economics, Henan University, Kaifeng, China

November 15, 2015

Microeconomic Theory

Week 9: Topics in Consumer Theory (Jehle and Reny, Chapter

November
2)
15, 2015

1 / 29

2.1 Duality: A Closer Look

The solutions to utility maximization problems and expenditure
minimization problems are, in sense, the same. In this section, we
shall explore further the connections among direct utility, indirect utility
(from the maximization problem) and expenditure functions (from the
minimization problem). The first duality is that every function of prices
and utility that has all the properties of an expenditure function is in
fact an expenditure function, i.e. there is a well-behaved utility function
that generates it.
Consider any function of prices and utility, E(p, u), that may or may not
be an expenditure function. Suppose this function satisfies the
expenditure function properties 1 to 7 of Theorem 1.7, so that it is
continuous, strictly increasing, and unbounded above in u, as well as
increasing in, homogeneous of degree one, concave, and differentiable
in p. To show E is exactly an expenditure function, it must be shown
there must exist a utility function on Rn+ whose expenditure function is
precisely E. Indeed we shall give an explict procedure for constructing
Microeconomic Theory

Week 9: Topics in Consumer Theory (Jehle and Reny, Chapter

November
2)
15, 2015

2 / 29

this utility function. To see how the construction works, choose

(p0 , u 0 ) Rn++ R+ , and evaluate E to obtain a number. Use this
number to construct the "half-space" in the consumption set,
A(p0 , u 0 ) {x Rn+ |p0 x E(p0 , u 0 )},
illustrated in Figure 2.1(a) (see the next slide). Now choose different
prices p1 , keep u 0 fixed, and construct another set,
A(p1 , u 0 ) {x Rn+ |p1 x E(p1 , u 0 )}.
Imagine proceeding like this for all prices p  0 and forming the
infinite intersection
A(u 0 ) p0 A(p, u 0 ) = {x Rn+ |p x E(p, u 0 ) for all p  0}.(2.1)
The shaded area in Figure 2.1(b) illustrated the infinite intersection. It
is easy to imagine that as more and more prices are considered and
more sets are added to the intersection, the shaded area will be more
closely resemble a superior set for some quasiconcave real-valued
function.
Microeconomic Theory

Week 9: Topics in Consumer Theory (Jehle and Reny, Chapter

November
2)
15, 2015

3 / 29

Figure 2.1. (a) The closed half-space A(p0 , u 0 ).

(b) The intersection of a finite collection of the sets A(p0 , u 0 ).

Microeconomic Theory

Week 9: Topics in Consumer Theory (Jehle and Reny, Chapter

November
2)
15, 2015

4 / 29

The following theorem says the the function constructed by this way is
increasing, unbounded above and quasiconcave, the properties like a
direct utility fuunction.

Theorem 2.1 Constructing a Utility Function from an Expenditure

Function
Let E : Rn++ R+ R+ satisfy properties 1 through 7 of an
expenditure function given in Theorem 1.7. Let A(u) be as in (2.1).
Then the function u : Rn+ R+ given by
u(x) max{u 0|x A(u)}
is increasing, unbounded above, and quasiconcave. (This function,
u(x), represents convex, monotonic prefernces.)
If E(p, u) is really an expenditure function generated by some utility
function u(x). From the definition of an expenditure function,
p x E(p, u(x)) for all prices p  0. Because E is strictly increasing
in u, u(x) is the largest value of u such that p x E(p, u) for all prices
p  0. That is, u(x) is the largest value of u such that x A(u).
Microeconomic Theory

Week 9: Topics in Consumer Theory (Jehle and Reny, Chapter

November
2)
15, 2015

5 / 29

Theorem 2.1 tells us we can begin with an expenditure function and

use it to construct a direct utility function representing some convex,
monotonic preferences.The following theorem says a function
satisfying Theorem 1.7 is indeed an expenditure function.

Theorem 2.2 The Expenditure Function of Derived Utility, u, Is E

Let E(p, u), defined on Rn++ Rn+ , 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 nonnegative prices and utility,
E(p, u) = minx p x

s.t. u(x) u.

That is, E(p, u) is the expenditure function generated by derived utility

u(x).
Once we obtain the utitlity through an expenditure function, if the
underlying preferences are continuous and strictly increasing, we can
invert the function in u, obtain the associated indirect utility function,
apply Roys identity, and derive the system of Marshallian demands as
well.
Microeconomic Theory

Week 9: Topics in Consumer Theory (Jehle and Reny, Chapter

November
2)
15, 2015

6 / 29

What if the utiltiy function is not strictly increasing and not

quasiconcave? What would be the properties of derived utility function
constructed from an expenditure function? Suppose only that u(x) is
continuous. Let e(p, u) be the expenditure function generated by u(x).
The continuity of u(x) is enough to guarantee that e(p, u) is
well-defined and continuous. Consider the utility function, call it w(x),
generated by e() in the now familiar way, that is,
w(x) max{u 0|p x e(p, u) p  0}
By Theorem 2.1, w(x) is increasing and quasiconcave. Thus,
regardless of whether or not u(x) is quasiconcave or increasing, w(x)
will be both increasing and quasiconcave. Clearly, u(x) and w(x) need
not coincide.
By the definition of e(p, u), that is, p x e(p, u), and the defintion of
w(x), we can see w(x) u(x). For any number u 0, the level-u
superior set for u(x), say S(u), will be contained in the level-u superior
set for w(x), say T (u). Moreover, because w(x) is quasiconcave, T (u)
is convex.
Microeconomic Theory

Week 9: Topics in Consumer Theory (Jehle and Reny, Chapter

November
2)
15, 2015

7 / 29

Figure 2.2. Duality between expenditure and utility.

Microeconomic Theory

Week 9: Topics in Consumer Theory (Jehle and Reny, Chapter

November
2)
15, 2015

8 / 29

In Figure 2.2 (a), u(x) is quasiconcave and increasing, then the

boundary of S(u) yields the negatively sloped, convex indifference
curve u(x) = u. Note that each point on the boundary is the
expenditure-minimizing bundle to achieve utility u at some price vector
p  0. For any number u, by the definition of the expenditure function
it must be that u(x) u, which implies u(x) u w(x). But it had
been said w(x) u(x). Therefore, w(x) = u(x).
The case depicted in Figure 2.2 (b) is more interesting. There, u(x) is
neither increasing nor quasiconcave. Note that some bundles on the
indifference curve never minimize the expenditure required to obtain
utility level u regardless of the price vector. The thick lines in Figure
2.2(c) show those bundles that do minimize expenditure at some
positive price vector. For those bundle x on the thick linesegments in
Figure 2.2(c), we therefore have as before that w(x) = u(x) = u. But
because w(x) is quasiconcave and increasing, the w(x) = u
indifference curve must be as depicted in Figure 2.2(d). Thus, w(x)
differs from u(x) only as much as is required to become strictly
increasing and quasiconcave.
Microeconomic Theory

Week 9: Topics in Consumer Theory (Jehle and Reny, Chapter

November
2)
15, 2015

9 / 29

The implication for this case is that any observable demand behavior
that can be generated by a nonincreasing, nonquasiconcave utility
function, like u(x), can also be generated by an increasing,
quasiconcave utility function, like w(x). It is in the sense that the
assumptions of monotonicity and convexity of preferences have no
observable implications for our theory of consumer demand.
Because the expenditure function and indirect utility are closely related
(i.e. are inverses of each other), the duality between the expenditure
function and the direct utility function implies there exists a duality
between the indirect utility function and direct utility function. Suppose
that u(x) generates the indirect utility function v (p, y ). Then by
definition of the indirect utility function (see (1.12)), for every x Rn+ ,
v (p, p x) u(x) holds for every p  0. In addition, there will be be
some price vector for which the equality holds. Evidently, we may write
u(x) = minpRn++ v (p, p x). (2.2)
(2.2) provides a mean for recovering the utility function u(x) from
knowledge of only the indirect utility function it generates.
Microeconomic Theory

Week 9: Topics in Consumer Theory (Jehle and Reny, Chapter

November
2) 15, 2015

10 / 29

Theorem 2.3 Duality Between Direct and Indirect Utility

Suppose that u(x) is quasiconcave and differentiable on Rn++ with
strictly positive partial derivative there. Then for all x Rn++ , v (p, p x),
the indirect utility function generated by u(x), achieves a minimum in p
on Rn++ , and
u(x) = minpRn++ v (p, p x). (T.1)
Beacuse v (p, y ) is homogeneous of degree zero in (p, y ), we have
v (p, p x) = v (p/(p x), 1) whenever p x > 0. Consequently, let
p x = 1 we can rewrite (T.1) as
u(x) = minpRn++ v (p, 1). s.t. p x = 1 (T.1)
Whether we use (T.1) or (T.1) to recover u(x) from v (p, y ) does not
matter. Simply choose that which is more convenient. One
disadvantage of (T.1) is that it always possess multiple solutions
becuase of the homogeneity of v (i.e., if p solves (T.1), then so does
tp for all t > 0).
Microeconomic Theory

Week 9: Topics in Consumer Theory (Jehle and Reny, Chapter

November
2) 15, 2015

11 / 29

Example 2.1
Convert the indirect utility function v (p, y ) = y (p1r + p2r )1/r to the
direct utility function. We use (T.1) by setting y = 1, which yields
v (p, 1) = (p1r + p2r )1/r . The direct utility function therefore will be the
minimum-value function
u(x1 , x2 ) = minp1 ,p2 (p1r + p2r )1/r s.t. p1 x1 + p2 x2 = 1.
The first order conditions for the Lagrangian require that the optimal p1
and p2 satisfy
(p1r + p2r )(1/r )1 (p1 )r 1 x1 = 0,
(p1r + p2r )(1/r )1 (p2 )r 1 x2 = 0,
1 p1 x1 p2 x2 = 0.
Solve the system of equations to obtain
p1 =
Microeconomic Theory

1/(r 1)
x1
r /(r 1)
r /(r 1)
x1
+x2

, p2 =

1/(r 1)
x2
r /(r 1)
r /(r 1)
x1
+x2

Week 9: Topics in Consumer Theory (Jehle and Reny, Chapter

November
2) 15, 2015

12 / 29

Example 2.1 (Continued)

Substituting these into the objective function and forming u(x1 , x2 ), we
obtain
" r /(r 1)
#1/r
r /(r 1)
h
i
x1
+ x2
r /(r 1)
r /(r 1) 1r 1/r
u(x1 , x2 ) =
=
(x
+
x
)
1
2
r /(r 1)
r /(r 1) r
(x1
+ x2
)
r /(r 1)

= (x1

r /(r 1) (r 1)/r

+ x2

Defining r /(r 1) yields

u(x1 , x2 ) = (x1 + x2 )1/
This is the CES direct utility function we started with in Example 1.2,
as it should be.
The last duality result we take up concerns the consumers inverse
demand function. The following theorem shows how to obtain the
inverse demand from the direct utility function without solving the
maximization problem.
Microeconomic Theory

Week 9: Topics in Consumer Theory (Jehle and Reny, Chapter

November
2) 15, 2015

13 / 29

Theorem 2.4 (Hoteling, Wold) Duality and the System of Inverse

Demands
Let u(x) be the consumers direct utility function and it differentiable.
Then the inverse demand function for good i associated with income
y = 1 is given by
u(x)/xi
pi (x) = Pn
j=1 xj (u(x)/xj )

Example 2.2
Lets take the case of the CES utility function once again. If
u(x1 , x2 ) = (x1 + x2 )1/ , then
u(x)/xj = (x1 + x2 )(1/)1 xj1
multiplying by xj , summing over j = 1, 2, forming the required ratios,
and invoking Theorem 2.4 gives the following system of inverse
functions when income y = 1:
p1 = x11 (x1 + x2 )1 ; p2 = x21 (x1 + x2 )1
Microeconomic Theory

Week 9: Topics in Consumer Theory (Jehle and Reny, Chapter

November
2) 15, 2015

14 / 29

2.2 Integrability
In Chapter 1.5, we showed that a utility maximizing consumers
demand function must satisfy homogeneity of degree zero in prices
and income, budget balancedness, symmetric, and negative
semidefiniteness, along with Cournot and Engel aggregation. But from
Theorem 1,17 the two aggregation results are derived from budget
balancedness so they are redundant. There is another redundancy as
well. Of the remaining four conditions, only budget balancedness,
symmetry, and negative semidefiniteness are truly independent:
Homogeneity of degree zero is implied by budget balancedness and
symmetry.

Theorem 2.5 Budget Balancedness and Symmetry Imply

Homogeneity
If x(p, y ) satisfies budget balancedness and its Slutsky matrix is
symmetric, then it is homogeneous of degree zero in p and y .
Proof: When budget balancedness holds,
Microeconomic Theory

Week 9: Topics in Consumer Theory (Jehle and Reny, Chapter

November
2) 15, 2015

15 / 29

Theorem 2.5 (Continued)

y=

Pn

i=1 pi xi (p, y )

Take the first derivatives with respect to prices and income to obtain
for, i = 1, , n,
n
X

pj

j=1

xj (p, y )
= xi (p, y ),
pi

n
X
j=1

pj

(P.1)

xj (p, y )
= 1. (P.2)
y

Fix p and y , then let fi (t) = xi (tp, ty ) for all t > 0. Differentiating fi with
respect to t gives
fi0 (t) =

n
X
xi (tp, ty )
j=1

Microeconomic Theory

pj

pj +

xi (tp, ty )
y (P.3)
y

Week 9: Topics in Consumer Theory (Jehle and Reny, Chapter

November
2) 15, 2015

16 / 29

Theorem 2.5 (Continued)

Now by budget balancedness, tp x(tp, ty ) = ty , so that dividing by
t > 0, we may write:
y=

Pn

j=1

pj xj (tp, ty ) (P.4)

Substituting from (P.4) for y in (P.3) and rearranging yields

fi0 (t)

n
X


pj

j=1


xi (tp, ty ) xi (tp, ty )
+
xj (tp, ty )
pj
y

The term in bracket is the ij th element of the Slutsky matrix. When

symmetry of the Slutsky matrix holds, the ij th element is equal to
the ji th element. Consequently we may interchange i and j within
those brackets and maintain equality. Therefore,
fi0 (t)

n
X
j=1

Microeconomic Theory


pj


xj (tp, ty ) xj (tp, ty )
+
xi (tp, ty )
pi
y

Week 9: Topics in Consumer Theory (Jehle and Reny, Chapter

November
2) 15, 2015

17 / 29

Theorem 2.5 (Continued)

n
n
X
X
xj (tp, ty )
xj (tp, ty )
=
pj
+ xi (tp, ty )
pj
pi
y
j=1
j=1

n
n
X
X
xj (tp, ty )
xj (tp, ty )
1
1
=
tpj
tpj
+ xi (tp, ty )
t
pi
t
y
j=1

j=1

1
1
[xi (tp, ty )](by (P.1)) + xi (tp, ty ) [1](by (P.2))
t
t
= 0.

Since the demand fi (t) = xi (tp, ty ) does not change in t, it is

homogeneous of degree zero.
In summary, for a utility-maximizers system, the demand function must
satisfies the three properties,
Microeconomic Theory

Week 9: Topics in Consumer Theory (Jehle and Reny, Chapter

November
2) 15, 2015

18 / 29

Budget Balancedness: p x(p, y ) = y

Negative Semidefiniteness: The associated Slutsky matrix s(p, y )
must be negative semidefinite.
Symmetry: s(p, y ) must be symmetric.
On the other hand, it can be proved that demand behavior is
consistent with the theory of utility maximization if and only if it satisfies
budget balancedness, negative semidefiniteness, and symmetry. This
impressive result warrants a formal statement.

Theorem 2.6 Integrability Theorem

n
A continuously differentiable function x : Rn+1
++ R++ is the demand
function generated by some increasing, quasiconcave utility function if
(and only if, when utility is continuous, strictly increasing, and strictly
quasiconcave) it satisfies budget balancedness, symmetry, and
negative semidefiniteness.

Microeconomic Theory

Week 9: Topics in Consumer Theory (Jehle and Reny, Chapter

November
2) 15, 2015

19 / 29

In practice, this theorem allows us to recover the utility function given

the demand functions satisfying the three properties.

Example 2.3
Suppose there are three goods and that a consumers demand
behavior is summarized by the functions
xi (p1 , p2 , p3 , y ) =

i y
,
pi

i = 1, 2, 3,

where i > 0, and 1 + 2 + 3 = 1. It is straightforward to check that

the demand functions satisfies budget balancedness, and the Slutsky
equation is symmetric and negative semidefinite. Consequently, by
Theorem 2.6, x(p, y ) must be utility-generated.
The first step is to derive the expenditure function. Remember the
Shephardd lemma
e(p1 , p2 , p3 , u)
= xih (p1 , p2 , p3 , u)
pi
Microeconomic Theory

Week 9: Topics in Consumer Theory (Jehle and Reny, Chapter

November
2) 15, 2015

20 / 29

Example 2.3 (Continued)

By duality between Hicksian demand functions and Marshallian
demand function,
xih (p1 , p2 , p3 , u) = xi (p1 , p2 , p3 , e(p1 , p2 , p3 , u)) =

i e(p1 , p2 , p3 , u)
pi

Therefore,
i e(p1 , p2 , p3 , u)
e(p1 , p2 , p3 , u)
=
,
pi
pi

i = 1, 2, 3

This is a partial differential equation. Move e(p, u) to the left hand side,
pi to the right hand side, and take the integral on the both side
Z
Z
e(p, u)
pi
= i
, i = 1, 2, 3
e(p, u)
pi
You would have no trouble deducing that
Microeconomic Theory

Week 9: Topics in Consumer Theory (Jehle and Reny, Chapter

November
2) 15, 2015

21 / 29

Example 2.3 (Continued)

ln(e(p, u)) = i ln(pi ) + constant,

i = 1, 2, 3

The only additional element to keep in mind is that when partially

differentiating with respect to, say, pi , all the other variables, p1 , p3 , and
u, are treated as constants. With this in mind, it is easy to see the
three equations above imply the following three:
ln(e(p, u)) = 1 ln(p1 ) + c1 (p2 , p3 , u),
ln(e(p, u)) = 2 ln(p2 ) + c2 (p1 , p3 , u),
ln(e(p, u)) = 3 ln(p3 ) + c3 (p1 , p2 , u),
where the ci () functions are the constant terms. Because the three
equations should hold simultaneously, the only possibility is
ln(e(p, u)) = 1 ln(p1 ) + 2 ln(p2 ) + 3 ln(p3 ) + c(u),
where c() is some function of u. But this means that
Microeconomic Theory

Week 9: Topics in Consumer Theory (Jehle and Reny, Chapter

November
2) 15, 2015

22 / 29

Example 2.3 (Continued)

e(p, u) = c(u)0 p11 p22 p33
Because e(p, u) is strictly increasing in u, we can choose any c(u)0
that is strictly increasing. It does not matter what strictly increasing
functions we choose because any monotonic transform can keep the
order of utility. We may choose c(u)0 = u, so that
e(p, u) = up11 p22 p33
We can then know the utility function.

Microeconomic Theory

Week 9: Topics in Consumer Theory (Jehle and Reny, Chapter

November
2) 15, 2015

23 / 29

2.3 Revealed Preference

We had discussed about the properties of demand functions derived
from the consumers preference or utility. The problem is that
preference is not observable so we cannot directly examine the
demand functions. The theory of revealed preference says that
virtually every prediction ordinary consumer theory makes for a
consumers observable market can also be derived from a few simple
and sensible assumption about the consumers observable choices
themselves, rather than about his unobservable preferences.
The basic idea is simple: If the consumer buys one bundle instead of
another affordable bundle, then the first bundle is considered to be
revealed preferred to the second. Thus his tastes had been revealed
by his choices. Instead of laying down axioms on a persons
preferences as well we did before, we instead make assumptions
about the consistency of the choices that are made. Formally,
Microeconomic Theory

Week 9: Topics in Consumer Theory (Jehle and Reny, Chapter

November
2) 15, 2015

24 / 29

Definition 2.1 Weak Axiom of Revealed Preference (WARP)

A consumers choice behavior satisfies WARP if for every distinct pair
of bundles x0 , x1 with x0 chosen at prices p0 and x1 chosen at prices
p1 ,
p0 x1 p0 x0 = p1 x0 > p1 x1
In other words, WARP holds if whenever x0 is revealed preferred to x1 ,
x1 is never revealed preferred to x0 .
To better understand the implications of this definition, look at Figure
2.3.(see the next slide). In both graphs, the consumer facing p0
chooses x0 , and facing p1 chooses x1 . In (a), when the price is at p0 ,
the consumer chooses x0 although x1 is affordable to him (x1 is inside
the budget constraint). When the prices change from p0 to p1 , he
chooses x1 because x0 becomes not affordable to him. Therefore, his
behavior does not violate WARP.
Microeconomic Theory

Week 9: Topics in Consumer Theory (Jehle and Reny, Chapter

November
2) 15, 2015

25 / 29

In (b), at price p0 , both x0 and x1 are affordable to him. His choice of

x0 means he prefers x0 to x1 . When the prices change from p0 to p1 ,
he chooses x1 instead of x0 although x0 is affordable to him.
Figure 2.3. The Weak Axiom of Revealed Preference (WARP)

Microeconomic Theory

Week 9: Topics in Consumer Theory (Jehle and Reny, Chapter

November
2) 15, 2015

26 / 29

A Numerical Example: WARP

In the table above, first bundle is chosen at first prices, second bundle
at second prices and third bundle at third prices. So the diagonal
corresponds to actual choices.
From the first row, because we see the consumer chooses (10, 1)
although (5, 5) and (5, 4) are affordable. So we know (10, 1) % (5, 5)
and (10, 1) % (5, 4). From the second row, we know (5, 5) % (5, 4).
From the third row, we know (5, 5) % (10, 1) which contradicts
(10, 1) % (5, 5). His behavior violates WARP.
Microeconomic Theory

Week 9: Topics in Consumer Theory (Jehle and Reny, Chapter

November
2) 15, 2015

27 / 29

Now suppose a consumers choice behavior satisfies WARP. Let

x(p, y ) denote the choice function (note well this is not a demand
function because we have not mentioned utility or utility maximization),
the quantities the consumer chooses facing prices p and income y .
For p  0, we assume the choice function x(p, y ) satisfies budget
balancedness, i.e. p x(p, y ) = y . The first consequence of WARP and
budget balancedness is that the choice function x(p, y ) must be
homogeneous of degree zero in (p, y ).
Next consequence of WARP is the negative semidefiniteness of the
Slutsky equation. Then it can also be proved the Slutsky equation is
symmetric by budget balancedness and homogeneity of degree zero.
However, symmetry only applies in the case of two goods. When there
are only two goods, the choice function x(p, y ) satisfies both negative
definiteness and symmetry given the assumptions of WARP and
budget balancedness. Then the choice function is actually a demand
function because we would then be able to construct a utility function
generating it. On the other hand, we also can use the choice function
to recover the utility function using integrability.
Microeconomic Theory

Week 9: Topics in Consumer Theory (Jehle and Reny, Chapter

November
2) 15, 2015

28 / 29

For more than two goods, WARP and budget balancedness does not
necessarily imply symmetry of the Slutsky matrix because of the lack
of transitivity. To have a choice function to be equivalent to the demand
functions, we need a stronger axiom than WARP imposed on the
consumers behavior, that is Strong Axiom of Revealed Preference
(SARP).

Definition: Strong Axiom of Revealed Preference (Mas-Colell et.

al, 1995)
The choice function x(p, y ) satisfies the strong axiom of revealed
preference if for any list,
(p1 , y 1 ), (p2 , y 2 ), , (pN , y N )
with xn+1 (pn+1 , y n+1 ) 6= xn (pn , y n ) for all n N 1, we have
pN x1 (p1 , y 1 ) > y N whenever pn xn+1 (pn+1 , y n+1 ) y n for all
n N 1.
In words, if x1 (p1 , y 1 ) is directly or indirectly preferred to xN (pN , y N ),
then xN (pN , y N ) cannot be revealed preferred to x1 (p1 , y 1 ).
Microeconomic Theory

Week 9: Topics in Consumer Theory (Jehle and Reny, Chapter

November
2) 15, 2015

29 / 29