Solution to Problem Set Six - Choice Over

Budget Sets and the Dual Problem

Lecture Notes in Microeconomic Theory by Ariel Rubinstein
Last Update: Oct 8, 2007

1. In a world with two commodities, consider a consumers preferences that

are represented by the utility function u(x) = min{x1 , x2 }.

(a) Calculate the consumer’s demand function.

The consumer will set x1 = x2 by the functional form of u(x) and set
p1 x1 + p2 x2 = w by monotonicity. Consequently, p1 x2 + p2 x2 = w,
w
which implies that x(p, w) = ( p1 +p , w ).
2 p1 +p2

(b) Verify that preferences satisfy convexity.

Let y, z % x. Then yi , zi ≥ min{x1 , x2 } for i = 1, 2, and thus αyi +
(1 − α)zi ≥ min{x1 , x2 } for any α ∈ (0, 1), i = 1, 2. Consequently,
αy + (1 − α)z % x.

(c) Calculate the indirect utility function v(p, w).

w


v(p, w) = u x(p, w) = p1 +p 2
.

(d) Verify Roy’s Identity.

∂v(p,w) −w
∂p
i (p1 +p2 )2 w
xi (p, w) = − ∂v(p,w) =− 1 = .
p1 +p2
p1 + p2
∂w

(e) Calculate the expenditure function e(p, u) and verify the Dual Roy’s
Identity.
The agent’s expenditure minimization problem is

min px s.t. u(x) ≥ u.

x∈<2+

It’s optimal for the agent to choose xi = u for i = 1, 2, and thus

e(p, u) = u(p1 +p2 ). To verify Roy’s Dual Identity, note that hi (p, u) =
∂e(p,u)
∂pi = u.

1
2. Imagine that you are reading a paper in which the author uses the indirect
utility function v(p1 , p2 , w) = pw1 + pw2 . You suspect that the authors con-
clusions in the paper are the outcome of the“fact” that the function v is
inconsistent with the model of the rational consumer. Take the following
steps to make sure that this is not the case:

(a) Use Roys Equality to derive the demand function.

Let i ∈ {1, 2} and j ∈ {1, 2} such that i 6= j. Then

∂v(p,w) −w
∂pi p2i wpj
xi (p, w) = − ∂v(p,w)
=− p1 +p2 = .
p1 p2
pi (p1 + p2 )
∂w

(b) Show that if demand is derived from a smooth utility function, √

then
x
the indifference curve at the point (x1 , x2 ) has the slope − √x21 .
Let (p, w) ∈ <3++ , and note from (a) that x(p, w) is in the interior of
B(p, w). If we further assume that u is quasi-concave, then
v
∂u(x) u wp1 r
∂x1 p1 u p2 (p1 +p2 ) x2
∂u(x)
= = t
wp2 = .
p2 p1 (p1 +p2 ) x1
∂x2

(c) Construct a utility function with the property √

that the ratio of the
x2
partial derivatives at the bundle (x1 , x2 ) is x1 .
√
√ √
2
Let u(x) = x1 + x2 , and note that u(x) satisfies the condition
for the ratio of the partials.

(d) Calculate the indirect utility function derived from this utility func-
tion. Do you arrive at the original v(p1 , p2 , w)? If not, can the
original indirect utility function still be derived from another utility
function satisfying the property in (c)?
Yes, u(x) corresponds to the indirect utility function:
r r 2
wp2 wp1
u(x(p, w)) = + =
p1 (p1 + p2 ) p2 (p1 + p2 )
 √ 2
w(p1 + p2 ) w(p1 + p2 )2 w w
p = = + = v(p, w).
p1 p2 (p1 + p2 ) p1 p2 (p1 + p2 ) p1 p2

2
3. A consumer with wealth w is interested in purchasing only one unit of
one of the items included in a (finite) set A. All items are indivisible.
The consumer does not derive any “utility” from leftover wealth. The
consumer evaluates commodity x ∈ A by the number Vx (where the value
of not purchasing any of the goods is 0). The price of commodity x ∈ A
is px > 0.

(a) Formulate the consumer’s problem.

Let d denote the consumer’s action of “not purchasing anything.”
Set Vd = 0 and pd = 0, ie “not purchasing anything” generates no
utility and is costless. Define the set A0 = A ∪ {d}. The problem is:

max Vx s.t. px ≤ w.
x∈A0

(b) Check the properties of the indirect utility function (homogeneity of

degree zero, monotonicity, continuity and quasi-convexity).
Define B(p, w) = {x ∈ A0 | px ≤ w}. Then

v(p, w) = max Vx .
x∈B(p,w)

Homogeneity of Degree Zero Yes. For all λ > 0, B(p, w) =

B(λp, λw), and thus

v(p, w) = max Vx = max Vx = v(λp, λw).

x∈B(p,w) x∈B(λp,λw)

Monotonicity Yes. Note that

p ≥ p0 , w ≤ w0 ⇒ B(p.w) ⊆ B(p0 , w0 ),

and thus v(p, w) ≤ v(p0 , w0 ).

Continuity No. Let A
= {x}, Vx = 1 and p = (pd , px ) = (0,
1).
Note that v (0, 1),
1 = 1 but for every  > 0, v (0, 1 + ), 1 = 0
and v (0, 1), 1 −  = 0. Thus v is not continuous in p or w.
Quasi-Convexity Yes. Let v(p, w) ≤ v(p0 , w0 ) and λ ∈ [0, 1], and
define (p00 , w00 ) = λ(p, w) + (1 − λ)(p0 , w0 ). To prove quasi-
convexity, we must show v(p00 , w00 ) ≤ v(p0 , w0 ).
Let x∗ be the best bundle in B(p00 , w00 ). Then

[λpx∗ + (1 − λ)p0x∗ ] = p00x∗ ≤ w00 = [λw + (1 − λ)w0 ], and thus

px∗ ≤ w or p0x∗ ≤ w0 .
x∗ is thus affordable in either B(p, w) or B(p0 , w0 ), and conse-
quently v(p00 , w00 ) ≤ v(p, w) or v(p00 , w00 ) ≤ v(p0 , w0 ), and thus in
either case v(p00 , w00 ) ≤ v(p0 , w0 ).

3
 (c) Calculate the indirect utility function for the case in which A = {a, b}
and Va > Vb > 0.

 Va if pa ≤ w
v(p, w) = Vb if pb ≤ w < pa
0 if pa , pb > w.


4. Let X be a set and % be preferences on X. Let D be a set of choice

problems and let %∗ be the indirect preference relation defined on D. One
route to elicit the choice function c% from %∗ is by concluding that:
c% (A) = x∗ when, for every y ∈ A \ {x∗ }, there is a set B(y) containing
x∗ but not y, such that B(y) %∗ A.
Explain this definition and explain the analogy to Roys equality.
Let x∗ ∈ A satisfy the condition described above. Then for every other
y ∈ A, there is a set containing x∗ and not y that is at least as good as
A. In other words, the %-maximal element in a set containing x∗ , but not
y, is at least as good as the %-maximal element in A. Since this holds for
all y ∈ A, then we can infer that x∗ is the %-maximal element in A.
For Roy’s Equality, we construct the set of price/wealth pairs such that
px(p∗ , w∗ ) = w. For each of these pairs, we have B(p, w) %∗ B(p∗ , w∗ )
since x(p, w) % x(p∗ , w∗ ). Consequently, given indirect preferences, we
can calculate the bundle x(p∗ , w∗ ) since the set of prices (p, w) is tangent
to the indifference set through the pair (p∗ , w∗ ).

5. A consumer holds continuous preference relation % (but the optimization

% over B(p, w) does not necessarily yield a unique solution). State and
prove the four properties of the induced indirect preferences %∗ which are
analogous to the four properties stated and proved for the case that x(p, w)
is always well defined.
Let x(p, w) denote the set of optimal bundles in B(p, w).

(a) (λp, λw) ∼∗ (p, w): Since x(p, w) is the set of optimal bundles in
B(p, w), then u(x) = u(x0 ) for all x, x0 ∈ x(p, w). Then (λp, λw) ∼∗
(p, w) because x(λp, λw) = x(p, w).
(b) %∗ is non-increasing in pk , increasing in w: Reducing the size of the
budget set cannot be beneficial. Moreover, if w increases, the agent
can consume more of all commodities.
(c) If % is continuous, then %∗ is continuous and there exists a continuous
v representing %∗ : Let u be a continuous utility function representing
%. Define v(p, w) = u(x∗ ), where x∗ is any element of x(p, w), and

4
 note that v is well-defined since all elements in x(p, w) yield the same
level of utility. Moreover, note that v represents %∗ .
First, let’s show v is continuous in p. By contradiction, assume that
there exists a convergent sequence of prices pn → p such that v(pn , w)
does not converge to v(p, w). Define the sequence {xn }, where xn ∈
x(pn , w) for all n. Let’s show that {xn } converges. Define m =
inf{pnk | k ∈ {1, ..., K} and n ∈ N}, and note {xn } is contained in
w K
the compact set [0, m ] . Therefore, without loss of generality, we
can assume that x → y. Since, by assumption, v(pn , w) does not
n

converge to v(p, w), then it must be that y ∈ / x(p, w). Note that
py = lim pn xn ≤ w, and thus x(p, w)  y. From Question 5 in PS 5,
we know that x(p, w) is a closed set, and consequently there exists an
 > 0 such that z − e  y for all z ∈ x(p, w) by continuity. But then
for n large enough, pn [z − e] ≤ w and z − e  xn for all z ∈ x(p, w),
a contradiction to xn ∈ x(pn , w). Then xn → x∗ ∈ x(p, w), and thus
u(xn ) → u(x∗ ) by the continuity of u. Consequently, v(pn , w) →
v(p, w).
Finally, take any convergent sequence (pn , wn ) → (p, w). Then
n
lim v(pn , wn ) = lim v( wp n , 1) = v( wp , 1) = v(p, w), where the first
and third equalities follow from (a) and the second equality follows
from v being continuous in p. Moreover, since v is continuous, then
%∗ is continuous as well.
(d) If (p1 , w1 ) %∗ (p2 , w2 ), then (p1 , w1 ) %∗ (p0 , w0 ), where p0 = λp1 +
(1 − λ)p2 and w0 = λw1 + (1 − λ)w2 for λ ∈ [0, 1]: Let z ∈ x(p0 , w0 ).
Then zp0 ≤ w0 , which implies that either p1 z ≤ w1 or p2 z ≤ w2 .
Therefore, z ∈ B(p1 , w1 ) or z ∈ B(p2 , w2 ). Since z was chosen arbi-
trarily, then either x(p1 , w1 ) % x(p0 , w0 ) or x(p2 , w2 ) % x(p0 , w0 ), and
since x(p1 , w1 ) % x(p2 , w2 ), then in either case x(p1 , w1 ) % x(p0 , w0 ).
Consequently, (p1 , w1 ) %∗ (p0 , w0 ).

6. Show that if the utility function is continuous, then so is the Hicksian

demand function h(p, u).
Let (pn , un ) → (p, u0 ) be a convergent sequence of price/utility pairs.
First, let’s show {hn } = {h(pn , un )} converges. Define u = sup{un } ∈
(0, ∞),
m = inf pni | n ∈ N and i ∈ {1, . . . , K} > 0,


M = sup pni | n ∈ N and i ∈ {1, . . . , K} < ∞,



pm = (m, ..., m) and pM = (M, . . . , M ). Let h = h(p, u), and note that
u(h) ≥ un for all n. Then

pm hn ≤ pn hn ≤ pn h ≤ pM h,

5
 where the first inequality follows from pm ≤ pn , the second inequality fol-
lows from hn being the cheapest bundle achieving utility un at prices pn
and the third inequality follows from pn ≤ pM . Therefore, hni ∈ [0, Mm hi ]
for all n and i = 1, ..., K, and consequently we can assume that hn con-
verges to some h∗ .
To complete the proof, assume by contradiction that h∗ 6= h(p, u). Since
u(hn ) ≥ un for all n, then the continuity of u implies that u(h∗ ) ≥ u0 .
Then it must be that ph∗ > ph(p, u). Let z  h(p, u) be such that
ph∗ > pz, and note that u(z) > u0 by monotonicity. By the continuity of
u, then for sufficiently large n, we have u(z) ≥ un . Moreover, since pn → p
and hn → h∗ , then pn hn > pn z for sufficiently large n, a contradiction to
hn being the cheapest bundle achieving utility un at prices pn .

7. A commodity k is a Giffen if the demand for the k-th good, xk (p, w), is
increasing in pk . A commodity k is inferior if the demand for the com-
modity decreases in wealth. Show that if k is Giffen in some neighborhood
of (p, w), then k is inferior.
For this proof, I will assume that demand for xk is strictly positive in the
neighborhood of (p, w). This assumption is innocuous since if xk (p, w) = 0,
then for p0k < pk , we would have xk (p0 , w) = xk (p, w) = 0, and for p0k > pk ,
then the agent would still be able to afford his original bundle x(p, w).
By contradiction, assume there exists a (p, w) such that xk (p, w) is Giffen
in the neighborhood of (p, w), but xk (p, w) is not inferior. Since xk (p, w)
is a Giffen, then xk (p + ek , w) > xk (p, w) for some  > 0. For notational
ease, define x = x(p, w) and x0 = x(p + ek , w).
Define a new price/wealth pair (p + ek , w + xk ), and define x00 = x(p +
ek , w + xk ). Since the k-th good is not inferior, then x00k ≥ x0k > xk .
Moreover, by construction

[p + ek ]x = w + xk ⇒ x ∈ B(p + ek , w + xk ),

and since xk < x00k , note that

[p + ek ]x00 = w + xk ⇒ px00 = w + [xk − x00k ] < w ⇒ x00 ∈ B(p, w).

Thus x ∼ x00 by the WA. Note, however, that there exists a z  x00 such
that pz ≤ w. But then z  x00 ∼ x by monotonicity, which contradicts x
being the optimal bundle in B(p, w).

8. One way to compare budget sets is by using the relation %∗ as defined

in the text. According to this approach, the comparison between (p, w)
and (p0 , w) is made by comparing two numbers u(x(p, w)) and u(x(p0 , w)),
where u is a utility function defined on the space of the bundles. Following

6
are two other approaches for making such comparisons using “concrete
terms.”
Define:
CV (p, p0 , w) = w − e(p0 , u) = e(p, u) − e(p0 , u)
where u = u(x(p, w)). This is the answer to the question: What is the
change in wealth that would be equivalent, from the perspective of (p, w),
to the change in price vectors from p to p0 ?
Define:
EV (p, p0 , w) = e(p, u0 ) − w = e(p, u0 ) − e(p0 , u0 )
where u0 = u(p0 , w). This is the answer to the question: What is the
change in wealth that would be equivalent, from the perspective of (p0 , w),
to the change in price vectors from p to p0 ?
Now, answer the following questions regarding a consumer in a two-commodity
world with a utility function u:

(a) For the case u(x1 , x2 ) = x1 +x2 , calculate the two “consumer surplus”
measures.
Let p, p0 be two price vectors. Then u = min{pw1 ,p2 } and u0 = min{pw0 ,p0 } .
1 2
From here, a bit of algebra yields
w min{p01 , p02 } min{p1 , p2 } − min{p01 , p02 }
 
CV (p, p0 , w) = w − =w
min{p1 , p2 } min{p1 , p2 }

min{p1 , p2 } − min{p01 , p02 }

 
0 w min{p1 , p2 }
EV (p, p , w) = −w =w
min{p01 , p02 } min{p01 , p02 }
6 min{p01 , p02 }.
which are different if min{p1 , p2 } =

(b) Explain why the two measures may give different values for some
other utility functions.
When a price changes, the consumer is effected in two ways: a direct
“price” effect and an indirect “wealth” effect. If the wealth effect is
different at the two bundles x(p, w) and x(p0 , w), then the CV and
EV will generate different values.

(c) Explain why the two measures are identical if the individual has quasi-
linear preferences in the second commodity and in a domain where
the two commodities are consumed in positive quantities.
Assume that preferences are also strictly monotonic in the second
commodity (in addition to the usual assumption of continuity and
monotonicity). Under this assumption, we proved in Chapter 4 that
preferences can be represented by u(x) = x2 +φ(x1 ). Moreover, recall
that when preferences are quasi-linear, a change in wealth causes a
change in demand for the quasi-linear commodity only, ie demand for
x1 is independent of wealth (we proved this in Question 4, PS5).

7
 Normalize prices such that p2 = p02 = 1, and note
x2 (p, e(p, u))+φ x1 (p, e(p, u)) = u = x2 (p0 , e(p0 , u))+φ x1 (p0 , e(p0 , u))

x2 (p, e(p, u0 ))+φ x1 (p, e(p, u0 )) = u0 = x2 (p0 , e(p0 , u0 ))+φ x1 (p0 , e(p0 , u0 )) .

Since x1 is independent of wealth, then x1 (p, e(p, u)) = x1 (p, e(p, u0 ))

and x1 (p0 , e(p0 , u)) = x1 (p0 , e(p0 , u0 )). By subtracting the above two
equations, we have:
x2 (p, e(p, u)) − x2 (p, e(p, u0 )) = x2 (p0 , e(p0 , u)) − x2 (p0 , e(p0 , u0 )).
Since p2 = p02 = 1, then x2 (p, u) = e(p, u) − p1 x1 (p, u) by the budget
constraint, and likewise for the other three terms. Substituting into
the above equation, it follows
[e(p, u) − p1 x1 p, e(p, u) ] − [e(p, u0 ) − p1 x1 p, e(p, u0 ) ] =

[e(p0 , u) − p01 x1 p0 , e(p0 , u) ] − [e(p0 , u0 ) − p01 x1 p0 , e(p0 , u0 ) ].

Again, since x1 is independent of wealth, all of the x1 terms cancel

out. After rearranging the terms, it follows
CV (p, p0 , w) = e(p, u) − e(p0 , u) = e(p, u0 ) − e(p0 , u0 ) = EV (p, p0 , w).

(d) Assume that the price of the second commodity is fixed and that the
price vectors differ only in the price of the first commodity. What
is the relation of the two measures to the “area below the demand
function” (which is a standard third definition of consumer surplus)?
Fix p2 , and let p01 < p1 , which implies that u = v(p, w) < v(p0 , w) =
u0 (if, as in (c), we assume x1 (p, w) > 0). For the time being, assume
that commodity 1 is a normal good, ie increasing in u. Then
h1 ((p1 , p2 ), u) ≤ h1 ((p1 , p2 ), v((p1 , p2 ), w))
≤ h1 ((p1 , p2 ), u0 ) for all p1 ∈ [p01 , p1 ].
It follows that
Z p1 Z p1
0 ∂e((p1 , p2 ), u)
CV (p, p , w) = dp1 = h1 ((p1 , p2 ), u)dp1
p01 ∂p1 p01
Z p1
≤ h1 ((p1 , p2 ), v((p1 , p2 ), w))dp1 ≤
p01
p1 p1
∂e((p1 , p2 ), u0 )
Z Z
h1 ((p1 , p2 ), u0 )dp1 = dp1 = EV (p, p0 , w).
p01 p01 ∂p1
Let A denote the area under the curve. Then CV ≤ A ≤ EV when
the good is normal. For inferior goods, EV ≤ A ≤ CV , since h1 (p, u)
would be decreasing in u. Finally, the relationship holds with equality
if the good is neither normal nor inferior, as discussed in part (b).