ECON 5113 Advanced Microeconomics
Winter 2019
Answers to Selected Exercises
The following questions are taken from Geoffrey A. Jehle
and Philip J. Reny (2011) Advanced Microeconomic The-
ory, Third Edition, Harlow: Pearson Education Limited.
The updated version is available at the course web page:
http://flash.lakeheadu.ca/∼kyu/E5113/Main.html
Ex. 1.11 Suppose that p ≥ 0 is a limit point of A.
Then for every  > 0, there exists a point q 6= p in
(p − , p + ) such that q ∈ A. This means that for ev-
ery neighbourhood B (pe) ∈ Rn+ , there is a bundle qe in
B (pe) ∩ % (x). Hence pe is a limit point of % (x). Since
% is continuous so that % (x) is a closed set, pe ∈ % (x),
which implies that p ∈ A. Therefore A is closed. The
proof for set B is similar.
Ex. 1.14 Let U be a continuous utility function that
represents %. Then for all x, y ∈ Rn+ , x % y if and only
if U (x) ≥ U (y).
First, suppose x, y ∈ Rn+ . Then U (x) ≥ U (y) or
U (y) ≥ U (x), which means that x % y or y % x. There-
fore % is complete.
Second, suppose x % y and y % z. Then U (x) ≥ U (y)
and U (y) ≥ U (z). This implies that U (x) ≥ U (z) and
so x % z, which shows that % is transitive.
Finally, let x ∈ Rn+ and U (x) = u. Then
U −1 ([u, ∞)) = {z ∈ Rn+ : U (z) ≥ u}
= {z ∈ Rn+ : z % x}
= % (x).
Since [u, ∞) is closed and U is continuous, % (x) is closed.
Similarly (I suggest you to try this), - (x) is also closed.
This shows that % is continuous.
Ex. 1.17 Suppose that a and b are two distinct bundle
such that a ∼ b. Let
A = {x ∈ Rn+ : αa + (1 − α)b, 0 ≤ α ≤ 1}.
and suppose that for all x ∈ A, x ∼ a. Then % is convex
but not strictly convex. Theorem 1.1 does not require
Instructor: Kam Yu
% to be convex or strictly convex, therefore the utility
function exists. Moreover, since % (a) = % (b) is convex,
there exists a supporting hyperplane H = {x ∈ Rn+ :
pT x = y} such that a, b ∈ H. Since H is an affine set,
A ⊂ H. This means that every bundle in A is a solution
to the utility maximization problem.
Ex. 1.341 Suppose on the contrary that E is bounded
above in u, that is, for some p  0, there exists M > 0
such that M ≥ E(p, u) for all u in the domain of E.
Let u∗ = V (p, M ). Then
E(p, u∗ ) = E(p, V (p, M )) = M = pT x∗ ,
where x∗ is the optimal bundle. Since U is continuous,
there exists a bundle x0 in the neighbourhood of x∗ such
that U (x0 ) = u0 > u∗ . Since U strictly increasing, E is
strictly increasing in u, so that E(p, u0 ) > E(p, u∗ ) =
M . This contradicts the assumption that M is an upper
bound.
Ex. 1.37 (a) Since x0 is the solution of the expenditure
minimization problem when the price is p0 and utility
level u0 , it must satisfy the constraint U (x0 ) ≥ u0 . Now
by definition E(p, u0 ) is the minimized expenditure when
price is p, it must be less than or equal to pT x0 since
x0 is in the feasible set, and by definition equal when
p = p0 .
(b) Since f (p) ≤ 0 for all p  0 and f (p0 ) = 0, it
must attain its maximum value at p = p0 .
(c) ∇f (p0 ) = 0.
(d) We have
∇f (p0 ) = ∇p E(p0 , u0 ) − x0 = 0,
which gives Shephard’s lemma.
Ex. 1.46 Since di is homogeneous of degree zero in p
and y, for any α > 0 and for i = 1, . . . , n,
di (αp, αy) = di (p, y).
1 It may be helpful to review the proof of Theorem 1.8.
Differentiate both sides with respect to α, we have (a) The Lagrangian is
n
∂di (αp, αy) Y
∇p di (αp, αy)T p + y = 0. L=A xα T
i − λ(p x − y).
i

∂y i=1

Put α = 1 and rewrite the dot product in summation The necessary conditions for maximization are the bud-
form, the above equation becomes get constraint and
Qn
n ∂L xα i
X ∂di (p, y) ∂di (p, y) = αj A i=1 i − λpj = 0,
pj + y = 0. (1) ∂xj xj
j=1
∂pj ∂y
for j = 1, . . . , n. Consider two goods i and j, the above
Dividing each term by di (p, y) yields the result. necessary condition implies that
αi /xi pi
Ex. 1.47 Suppose that U (x) is a linearly homogeneous = ,
αj /xj pj
utility function.
(a) Then which can be rearranged to
  
αj pi
E(p, u) = min{pT x : U (x) ≥ u} xj = xi .
x αi pj
T
= min{up x/u : U (x/u) ≥ 1}
x Substitute this relation for j = 1, . . . , n in the budget
= u min{pT x/u : U (x/u) ≥ 1} constraint, we have
x      
= u min{pT x/u : U (x/u) ≥ 1} (2) p1 α1 pi
xi +· · ·+pi xi +· · ·+pn
αn pi
xi = y,
x/u αi p1 αi pn
= u min{pT z : U (z) ≥ 1} (3) or
z   X n
!
= uE(p, 1) pi
αk xi = y.
αi
= ue(p) k=1
Pn
Since k=1 αk = 1, the above equation gives the Mar-
In (2) above it does not matter if we choose x or x/u shallian demand function of good i as
directly as long as the objective function and the con-
αi y
straint remain the same. We can do this because of the di (p, y) = xi = ,
objective function is linear in x. In (3) we simply rewrite pi
x/u as z. for i = 1, . . . , n.
(b) Using the duality relation between V and E and
the result from Part (a) we have Ex. 1.65 A homothetic preference relation means that
the utility function can be expressed as U (x) = g(f (x)),
y = E(p, V (p, y)) = V (p, y)e(p) where f : Rn+ → R+ is a linearly homogeneous function
and g : R+ → R+ is an increasing function. The ordinary
so that demand function is
y
V (p, y) = = v(p)y,
e(p) d(p, y) = argmax{U (x) : pT x = y}
x
where we have let v(p) = 1/e(p). The marginal utility
= argmax{g(f (x)) : pT x = y}
of income is x
∂V (p, y)
= v(p), = argmax{f (x) : pT x = y} (4)
∂y x

which depends on p but not on y. = argmax{yf (x/y) : pT x/y = 1} (5)

x

Ex. 1.54 The utility maximization problem is = argmax{f (x/y) : pT x/y = 1} (6)
x

n
Y = y argmax{f (x/y) : pT x/y = 1} (7)
max A xα
i
i x/y
x
i=1 = y argmax{f (z) : pT z = 1} (8)
T z
subject to p x = y,
= yd(p, 1)
Pn ˆ
where A > 0 and i=1 αi = 1. = y d(p). (9)

2
Equation (4) follows from the fact that an increasing Then
transform does not change the choice of the optimal bun-
yp1 log p2 yp2 log p1
dle x. Equation (5) uses the homogeneity property of p · d(p, y) = +
f . The budget constraint is divided by y. Again in (6) p1 log p2 + p2 log p1 p1 log p2 + p2 log p1
= y
yf (x/y)) is an increasing transform of f (x/y) and so the
optimal bundle remains the same without y. In equa- and therefore satisfies budget balancedness. It is straight
tion (7) we change the control variable from x to x/y so forward to verify that d(p, y) is not a homogenous func-
that we have to multiple the optimal bundle by y. The tion.
control variable x/y is written as z in (8).
Since d is homogeneous of degree zero in p and y, equa- Ex. 2.2 For i = 1, . . . , n, the i-th row of the matrix
ˆ
tion (9) implies that d(p) is homogeneous of degree −1. multiplication S(p, y)p is

Ex. 1.66 (b) By definition y 0 = E(p0 , u0 ), Therefore n 

X ∂di (p, y) ∂di (p, y)

pj + pj dj (p, y)
∂pj ∂y
y1 E(p1 , u0 ) j=i
> n n
y0 E(p0 , u0 ) X ∂di (p, y) ∂di (p, y) X
= pj + pj dj (p, y)
∂pj ∂y
means that y 1 > E(p1 , u0 ). Since the indirect utility j=i j=i
function V is increasing in income y, it follows that n
X ∂di (p, y) ∂di (p, y)
= pj + y (11)
1 1 1 1 1 0 0 ∂pj ∂y
u = V (p , y ) > V (p , E(p , u )) = u . j=i

=0 (12)
Ex. 1.67 It is straight forward to derive the expenditure where in (11) we have used the budget balancedness and
function, which is (12) holds because of homogeneity and (1) in Ex. 1.46.
p22
E(p, u) = p2 u − . (10) Ex. 2.3 By (T.1’) on p. 82,
4p1
U (x) = min {V (p, 1) : p · x = 1} .
(a) For p0 = (1, 2) and y 0 = 10, we can use (10) to n
p∈R++
obtain u0 = 11/2. Therefore, with p1 = (2, 1),
The Lagrangian is
u0 − 1/8 43 β
I= 0
= . L = −pα
1 p2 − λ(1 − p1 x1 − p2 x2 ),
2u − 1 80
(b) It is clear from part (a) that I depends on u0 . with the first-order conditions
(c) Using the technique similar to Exercise 1.47, it can
be shown that if U is homothetic, E(p, u) = e(p)g(u), −αpα−1
1 pβ2 + λx1 = 0
where g is an increasing function. Then
and
β−1
1 0
e(p )g(u ) 1
e(p ) −βpα 1 p2 + λx2 = 0.
I= 0 0
= 0
,
e(p )g(u ) e(p ) Eliminating λ from the first-order conditions gives
which means that I is independent of the reference utility β x1
p2 = p1 .
level. α x2

Ex. 2.1 Consider the case of one good. Let the demand Substitute this p2 into the constraint equation, we get
function be √ α 1
y p1 = ,
d(p, y) = √ . α + β x1
p
It is homogenous of degree zero but it does not satisfy and
β 1
budget balancedness. p2 = .
Conversely, consider the two-good case that the de- α + β x2
mand function is given by The utility function is therefore

αα β β
   
y log p2 y log p1 −β
d(p, y) = , . U (x) = x−α
1 x2 ,
p1 log p2 + p2 log p1 p1 log p2 + p2 log p1 (α + β)α+β

3
which is a Cobb-Douglas function. which is the CES function with ρ = 1/2. You should
verify with Example 1.3 on p. 39–41 that the expenditure
Ex. 2.6 We want to maximize utility u subject to the function is indeed as given.
constraint pT x ≥ E(p, u) for all p ∈ Rn++ . That is,
Ex. 3.2 Constant returns-to-scale means that f is lin-
up1 p2
p1 x1 + p2 x2 ≥ . early homogeneous. So by Euler’s theorem
p1 + p2
x1 ∂y/∂x1 + x2 ∂y/∂x2 = y. (15)
Rearranging gives
p1 + p2 p1 + p2 Since average product y/x1 is rising, its derivative respect
u≤ x1 + x2 to x1 is positive, that is,
p2 p1
for all p ∈ Rn++ . This implies that (x1 ∂y/∂x1 − y)/x21 > 0.

From (15) we have

 
p1 + p2 p1 + p2
u ≤ min x1 + x2 . (13)
p1 ,p2 p2 p1
x2 ∂y/∂x2 = −(x1 ∂y/∂x1 − y) < 0,
Therefore u attains its maximum value when equality
holds in (13). To find the minimum value on the right- which means that the marginal product ∂y/∂x2 is nega-
hand side of (13), write α = p2 /(p1 + p2 ) so that 1 − α = tive.
p1 /(p1 + p2 ) and 0 < α < 1. The minimization problem
becomes Ex. 4.1 If preferences are identical and homothetic, each
  consumer’s preferences can be represented by a linearly
x1 x2 homogeneous utility function. Using an argument sim-
min + :0<α<1 . (14)
α α 1−α ilar to exercise 1.47, the ordinary demand function of
consumer i is separable in p and yi , that is,
Notice that for any x1 > 0 and x2 > 0,
  di (p, yi ) = f (p)yi .
x1 x2
lim + =∞
α→0 α 1−α Market demand is therefore
and
X X
  di (p, yi ) = f (p) yi = f (p)y,
x1 x2
lim + =∞ i i
α→1 α 1−α P
so that the minimum value exists when 0 < α < 1. The where y = i yi is aggregate income. It is clear that
first-order condition for minimization is market demand depends on aggregate income y but not
on income distribution. The market level income elastic-
x1 x2 ity of demand for good j is
− + = 0,
α2 (1 − α)2
∂(fj (p)y) y
which can be written as ηj = = 1.
∂y fj (p)y
α2 x2 = (1 − α)2 x1 .
Ex. 4.2 If preferences are homothetic but not identical,
Taking the square root on both sides gives the demand function of consumer i is

αx2
1/2 1/2
= (1 − α)x1 . di (p, yi ) = f i (p)yi .

Rearranging gives Market demand is therefore

X X
x1
1/2
x2
1/2 di (p, yi ) = f i (p)yi .
α= 1/2 1/2
and 1−α= 1/2 1/2
. i i
x1 + x2 x1 + x2
In this case market demand depends on income distribu-
It is clear that α is indeed between 0 and 1. Putting tion.
α and 1 − α into the objective function in (14) give the
direct utility function Ex. 4.5 Let w be the vector of factor prices and p
 2 be the output price. Then the cost function of a typ-
1/2 1/2
U (x1 , x2 ) = x1 + x2 , ical firm with constant returns-to-scale technology is

4
C(w, y) = c(w)y where c is the unit cost function. The (c) If a > 0 and b > 0, the minimum efficiency scale
profit maximization problem can be written as is at q = 0. Therefore the long-run equilibrium market
price and number of firms are indeterminate.
max py − c(w)y = max y[p − c(w)].
y y
Ex. 4.12 In the Bertrand duopoly model of section 4.2.1,
For a competitive firm, as long as p > c(w), the firm will the firms have no fixed costs and equal marginal cost c.
increase output level y indefinitely. If p < c(w), profit In equilibrium both firms charge p = c and share the
is negative at any level of output except when y = 0. market equally.
If p = c(w), profit is zero at any level of output. In (a) Now suppose that firm 1 has fixed costs F > 0. If
fact, market price, average cost, and marginal cost are it stays in production, in equilibrium the price it charges
all equal so that the inverse supply function is a constant must be the same as firm 2. This implies that p1 = p2 =
function of y. Therefore the supply function of the firm p. But the zero profit condition for firm 1 is
does not exist and the number of firm is indeterminate.
pq1 − cq1 − F = 0,
Ex. 4.7 Suppose that all firms have the same technology
and therefore the same cost function. Given market price or F
p, profit of a typical firm j is pqj − c(qj ). p=c+ > c.
q1
(a) Suppose that a > 0, b < 0, and there are J firms in
the industry. The short-run profit maximization for firm Assume that the firm share the market equally, q1 = q2 =
j is Q/2 so that
J
!
max α − β
X
qi qj − (aqj + bqj2 ). 2F 2F
qj
p=c+ =c+ . (17)
i=1 Q α − βp
The necessary condition for profit maximization is Since firm 2 has no fixed cost, it can charge a price
J
slightly lower than this and capture the whole market.
(16) In equilibrium firm 2 charges a price that satisfies equa-
X
α − βqj − β qi − a − 2bqj = 0.
i=1
tion (17), which can be rearranged to the quadratic equa-
tion
By symmetry, q1 = q2 = · · · = qJ . Equation (16) be- βp2 − (α + βc)p + (αc + 2F ) = 0. (18)
comes
Therefore p2 = p, the solution in equation (18), q1 = 0,
α − β(J + 1)qj − a − 2bqj = 0,
and q2 = α − βp.
which gives (b) Suppose that both firms have fixed costs F > 0.
α−a Then both firms stay in production in equilibrium and
qj = .
β(J + 1) + 2b charge a price equal to the solution in equation (18).
The market output is They share the market with q1 = q2 = (α−βp)/2. Notice
that in equation (18), p = c is the solution if F = 0.
J(α − a) (c) If firm 1 has a lower marginal cost so that c2 >
q= ,
β(J + 1) + 2b c1 > 0, then firm 1 will charge a price p1 = c2 just to
keep firm 2 out of the market. Therefore q2 = 0 and
and the market price is q1 = α − βc2 .
βJ(α − a)
p=α− , Ex. 4.14 The profit maximization problem for a typical
β(J + 1) + 2b
firm is
(b) The average cost of production is
max [10 − 15q − (J − 1)q̄]q − (q 2 + 1),
q
c(q)
= a + bq,
q with necessary condition
which is a decreasing function if a > 0 and b < 0. Since 10 − 15q − (J − 1)q̄ − 15q − 2q = 0.
there is no fixed cost, a firm can potentially increase out-
put until the average (or total) cost of production is zero, (a) Since all firms are identical, by symmetry q = q̄.
which is at the output level q = −a/b. Notice that at zero This gives the Cournot equilibrium of each firm q ∗ =
market price, consumer demand is α/β. Therefore the 10/(J + 31), with market price p∗ = 170/(J + 31).
long-run equilibrium market price and number of firms (b) Short-run profit of each firm is π = [40/(J +31)]2 −
depend on the relative values of −a/b and α/β. 1. In the long-run π = 0 so that J = 9.

5
Ex. 4.19 The Marshallian demands are x∗ = 1/p and
m∗ = y − 1. The indirect utility function is V (p, y) =
y − log p − 1. If the price of x rises from p0 to p1 , the
compensating variation is implicitly defined as

V (p1 , y + CV) = V (p0 , y),

or
y + CV − log p1 − 1 = y − log p0 − 1.
This gives CV = log(p1 /p0 ), which is equal to the change
in consumer surplus.

Ex. 5.11 (a) The necessary condition for a Pareto-

efficient allocation is that the consumers’ MRS are equal.
Therefore Figure 1: Contract Curve and the Core
∂U 1 (x11 , x12 )/∂x11 ∂U 2 (x21 , x22 )/∂x21
1 1 1 = ,
1
∂U (x1 , x2 )/∂x2 ∂U 2 (x21 , x22 )/∂x22 Solving the quadratic equation gives one positive value
or of 14.16. Consumer 2’s utility function can be written
x12 x22 as x21 (x22 )2 . This can be expressed in terms of x11 and
= . (19)
x11 2x21 x12 using (20) and (21). The indifference curve passing
through endowment becomes
The feasibility conditions for the two goods are
(21 − x11 )(10 − x12 )2 = (21 − 18)(10 − 4)2 = 108.
x11 + x21 = e11 + e21 = 18 + 3 = 21, (20)
1
x12 + x22 = e12 + e22 = 4 + 6 = 10. (21) Putting x2 in (22) into the above equation and solving
for x1 give x11 = 15.21. Therefore the core of the economy
1
2 2 1 1
Express x1 in (20) and x2 in (21) in terms of x1 and x2 is given by
respectively, (19) becomes
10x11

C(e) = (x11 , x12 , x21 , x22 ) : x12 = ,
x12 10 − x12 42 − x11
= ,
x11 2(21 − x11 ) 14.16 ≤ x11 ≤ 15.21, x11 + x21 = 21,
or x12 + x22 = 10.

1
10x 1
x12 = . (22)
42 − x11 (c) Normalize the price of good 2 to p2 = 1. The
1
Eq. (22) with domain 0 ≤ x ≤ 21, (20), and (21) com- demand functions of the two consumers are:
1
pletely characterize the set of Pareto-efficient allocations y1 p1 e11 + p2 e12 18p1 + 4
A (contract curve). That is, x11 = = =
2p1 2p1 2p1

10x11 y1 1
p1 e1 + p2 e2 1
18p1 + 4
A = (x11 , x12 , x21 , x22 ) : x12 = , 0 ≤ x11 ≤ 21, x12 = = =
42 − x11 2p2 2p2 2
x11 + x21 = 21, x12 + x22 = 10.
y2 2
p1 e1 + p2 e2 2
3p1 + 6
x21 = = =
3p1 3p1 3p1
(b) The core is the section of the curve in (22) be- 2y 2 2 2
2(p1 e1 + p2 e2 ) 2(3p1 + 6)
tween the points of intersections with the consumers’ in- x22 = = =
3p2 3p2 3
difference curves passing through the endowment point.
For example, in Figure 1, if G is the endowment point, In equilibrium, excess demand z1 (p) for good 1 is zero.
the core is the portion of the contract curve between Therefore
points W and Z. Consumer 1’s indifference curve passing 18p1 + 4 3p1 + 6
through the endowment is + − 18 − 3 = 0,
2p1 3p1
(x11 x12 )2 = (18 · 4)2 , which gives p1 = 4/11 (check that market 2 also clears).
The Walrasian equilibrium is p = (p1 , p2 ) = (4/11, 1).
or x12 = 72/x11 . Substituting this into (22) and rearrang- From the demand functions above, the WEA is
ing give
5(x11 )2 + 36x11 − 1512 = 0. x = (x11 , x12 , x21 , x22 ) = (14.5, 5.27, 5.6, 4.73).

6
 (d) It is easy to verify that x ∈ C(e).
Ex. 5.23 Let Y ⊆ Rn be a strongly convex production
set. For any p ∈ Rn++ , let y1 ∈ Y and y2 ∈ Y be two
distinct profit-maximizing production plans. Therefore
p · y1 = p · y2 ≥ p · y for all y ∈ Y . Since Y is strongly
convex, there exists a ȳ ∈ Y such that for all t ∈ (0, 1),
1 2
ȳ > ty + (1 − t)y .
U (x) ≤ ū. Let u0 > ū and p0  0. Then by the
concavity of E in p, for any p  0,
E(p, u0 ) ≤ E(p0 , u0 ) + ∇p E(p0 , u0 )T (p − p0 )
= E(p0 , u0 ) + ∇p E(p0 , u0 )T p
−∇p E(p0 , u0 )T p0
= E(p0 , u0 ) + ∇p E(p0 , u0 )T p − E(p0 , u0 )
= ∇p E(p0 , u0 )T p,

Thus where in the second last equality we apply Euler’s theo-

rem to E since it is linearly homogeneous in p. Define
p · ȳ > tp · y1 + (1 − t)p · y2 x0 = ∇p E(p0 , u0 ) so that we have E(p, u0 ) ≤ pT x0 for
= tp · y1 + (1 − t)p · y1 all p  0. In other words, u0 is feasibility set of the
= p · y1 , maximization problem

which contradicts the assumption that y1 is profit- U (x0 ) = max{u : pT x0 ≥ E(p, u) ∀ p  0}.
maximizing. Therefore y1 = y2 .
Therefore U (x0 ) ≥ u0 , which contradicts the assumption
that ū is an upper bound of U .
Ex. 5.31 Let E = {(U i , ei , θij , Y j )|i ∈ I, j ∈ J } be
the production economy and p ∈ Rn++ be the Walrasian 2019 c The Pigman Inc. All Rights Reserved.
equilibrium.
(a) For any consumer i ∈ I, the utility maximization
problem is
X
max U i (x) s. t. p · x = p · ei + θij π j (p),
x
j∈J

with necessary condition

∇U i (x) = λp.

The MRS between two goods l and m is therefore

∂U i (x)/∂xl pl
i
= .
∂U (x)/∂xm pm
Since all consumers observe the same prices, the MRS is
the same for each consumer.
(b) Similar to part (a) by considering the profit maxi-
mization problem of any firm.
(c) This shows that the Walrasian equilibrium prices
play the key role in the functioning of a production econ-
omy. Exchanges are impersonal. Each consumer only
need to know her preferences and each firm its produc-
tion set. All agents in the economy observe the common
price signal and make their own decisions. This mini-
mal information requirement leads to the lowest possible
transaction costs of the economy.

Other Exercises
Theorem 2.1 Here is a suggested proof that the utility
function generated by the expenditure function is un-
bounded:
On the contrary suppose that U is bounded. That is,
there exists a utility level ū such that for all x ∈ Rn+ ,