ECO 503 Problem Set 4 Solutions

1. Here, y = f (x1 , x2 ) is a constant returns-to-scale production function. Suppose the average product of x1 is

strictly rising, that is, taking ω > 1, if we go from current level x1 to some ωx1 > x1 , the average product
f (ωx1 ,x2 ) f (x1 ,x2 )
at ωx1 will be strictly higher than at x1 , so ωx1 > x1 , or f (ωx1 , x2 ) > ωf (x1 , x2 ). Now, because

f has constant returns to scale, f (ωx1 , ωx2 ) = ωf (x1 , x2 ), which means f (ωx1 , x2 ) > f (ωx1 , ωx2 ). Notice

that choosing a higher input amount ωx2 > x2 results in a smaller output amount f (ωx1 , ωx2 ) < f (ωx1 , x2 ),

which means marginal product of x2 is negative. This problem is simple exercise that basically just requires

remembering the definitions.

2. Here, the production function f : RL→1

+ → R+ is continuous, strictly increasing and strictly concave, and

f (0) = 0, and (p, w) >> 0.

(a) The idea behind the proof is the following: When the price increases, even if the firm does not adjust its

action, the value of pf (x)↑w.x increases on its own, and if the firm then reoptimizes, it can only further

increase. We next present this argument in technical terms. We have to show that ε (p, w) = max pf (x)↑
x

w.x is increasing in p. Suppose that for any fixed w and for any p̂ˆ > p̂, x̂ ↓ arg max p̂f (x) →w .x. That
! x "
means ε (p̂, w) = p̂f (x̂) →w .x̂. Now, by definition of the profit function, ε p̂, ˆ w ↔ p̂fˆ (x̂) ↑ w.x̂ >
! "
ˆ w > ε (p̂, w).
p̂f (x̂) ↑ w.x̂ = ε (p̂, w). So we have shown that ε p̂,
! " ! "
(b) Now suppose for any fixed p and any ŵ ˆ > ŵ, x̂
ˆ ↓ arg max p̂f
ˆ (x) ↑ w.x. So ε p, ŵ ˆ = pf x̂ ˆ ↑ ŵ.
ˆ x̂.
ˆ
! " ! " x ! "
Now, by definition, ε (p, ŵ) ↔ pf x̂ ˆ ↑ ŵ.x̂
ˆ > pf x̂ˆ ↑ ŵ.
ˆ x̂
ˆ = ε p, ŵˆ .

(c) We have to now show that ε (p, w) is convex in (p, w). Take arbitrary (p̂, ŵ), (p̃, w̃) and ϑ ↓ [0, 1]. We

have to show ε (ϑp̂ + (1 ↑ ϑ) p̃, ϑŵ + (1 ↑ ϑ) w̃) ↗ ϑε (p̂, ŵ) + (1 ↑ ϑ) ε (p̃, w̃).

Let us define pε = ϑp̂ + (1 ↑ ϑ) p̃, wε = ϑŵ + (1 ↑ ϑ) w̃. Let xε ↓ arg max pε f (x) ↑ wε .x, so
x

ε (pε , wε ) = pε f (xε ) ↑ wε .xε .

1
 Now,

ε (pε , wε ) = pε f (xε ) ↑ wε .xε

= (ϑp̂ + (1 ↑ ϑ) p̃) f (xε ) ↑ (ϑŵ + (1 ↑ ϑ) w̃) .xε

= ϑ {p̂f (xε ) ↑ ŵ.xε } + (1 ↑ ϑ) {p̃f (xε ) ↑ w̃.xε }

↗ ϑε (p̂, ŵ) + (1 ↑ ϑ) ε (p̃, w̃)

↭ ε (ϑp̂ + (1 ↑ ϑ) p̃, ϑŵ + (1 ↑ ϑ) w̃) ↗ ϑε (p̂, ŵ) + (1 ↑ ϑ) ε (p̃, w̃)

The inequality above follows from the fact that when the prices are (p̂, ŵ) but the firm is still using the

input xε , it can only improve its profit by optimally choosing x for the changed prices. The idea behind

the overall result is that for some fixed x, the term pf (x) ↑ w.x linearly moves with (p, w), so without

adjusting x the firm can always get a linear change. But by optimizing, the changes will be better than

linear, that is, the profit function will always lie above the straight line. That makes it a convex function.

3. Consider the one-step profit maximization problem:

max pf (x) ↑ w.x

x

We can also write the choice variables separately, like so:

L→1
#
max pf (x1 , ..., xi , ..., xL→1 ) ↑ wj xj
x1 ,...,xi ,...,xL→1
j=1

ϑxi (p,w) ϑxi (p,w)

We are interested in the signs of ϑwi , ϑp and ϑy(p,w)
ϑp . By the MCS theorem, we know that the
2
$ %L→1 &
ϑxi (p,w) ϑ pf (x1 ,...,xi ,...,xL→1 )→ wj xj
(weak) sign of ϑwi is the same as the sign of ϑwi ϑxi
j=1
. Now,

' %L→1 (
ϖ 2 pf (x1 , ..., xi , ..., xL→1 ) ↑ j=1 wj xj ϖ
)
ϖf (x1 , ..., xi , ..., xL→1 )
*
= p ↑ wi
ϖwi ϖxi ϖwi ϖxi
= ↑1 ↗ 0

ϑxi (p,w)
Therefore, ↗ 0.
ϑwi $ %L→1 &
ϑxi (p,w) ϑ 2 pf (x1 ,...,xi ,...,xL→1 )→ wj xj
Similarly, for the sign of ϑp , we should find the sign of ϑpϑxi
j=1
. Now,

' %L→1 (
ϖ 2 pf (x1 , ..., xi , ..., xL→1 ) ↑ j=1 wj xj ϖ
)
ϖf (x1 , ..., xi , ..., xL→1 )
*
= p ↑ wi
ϖpϖxi ϖp ϖxi
ϖf (x1 , ..., xi , ..., xL→1 )
= ↔0
ϖxi

2
 ϑxi (p,w)
Therefore, ϑp ↔ 0.

Here, the inequality follows from the fact that f is an increasing function.

Lastly, for the sign of ϑy(p,w)

ϑp , notice that:

ϖy (p, w) ϖf (x (p, w))

=
ϖp ϖp
# ϖf (x (p, w)) ϖxi (p, w)
L→1
=
i=1 +
ϖxi ϖp
,- . + ,- .
↑0 ↑0
↔ 0

ϑf (x(p,w)) ϑxi (p,w) ϑf (x(p,w))

Here, we are summing the product ϑxi ϑp over all i ↓ {1, ..., L ↑ 1}. For each i, ϑxi ↔0
ϑxi (p,w)
because f is an increasing function, and ϑp ↔ 0 is what we showed a few lines ago. Both terms in the

product are positive, there each term in the sum is positive, and hence the sum is also positive.

4. Let f : RL→1
+ → R+ be a continuous and strictly increasing production function, with f (0) = 0, and let

C (w, y) be the associated cost function, where w ↓ RL→1

++ , that is, all input prices are strictly positive. We

have to show:

(a) C (w, y) = 0 when y = 0.

Because y = f (x) is strictly increasing and f (0) = 0, when y = 0 we must have x = 0. Then,

C (w, 0) = w.0 = 0.

(b) C (w, y) is strictly increasing in y.

Take arbitrary ŷ > ỹ and suppose x̂ ↓ z (w, ŷ) and x̃ ↓ z (w, ỹ), that is, x̂ and x̃ are the cost-

minimizing bundles for producing ŷ and ỹ, respectively. By definition, we must have f (x̂) ↔ ŷ and

f (x̃) ↔ ỹ. Because w >> 0, we must have f (x̂) = ŷ and f (x̃) = ỹ. Because f is strictly increasing and

ŷ = f (x̂) > f (x̃) = ỹ, we must have x̂ > x̃. This means, x̂i ↔ x̃i for all i ↓ {1, ..., L ↑ 1} and x̂i > x̃i

for some i ↓ {1, ..., L ↑ 1}. Now,

L→1
#
w.x̂ = wi x̂i
i=1
L→1
#
> wi x̃i
i=1
= w.x̃

The strict inequality comes from the fact that x̂i > x̃i for some i and wi > 0 for the same i, together

with the fact that x̂i ↔ x̃i for all i.

We have w.x̂ > w.x̃, so C (w, ŷ) > C (w, ỹ). Because ŷ and ỹ were arbitrary, we have proved that

C (w, y) is strictly increasing in y.

3
(c) C (w, y) is increasing in w.

Take arbitrary ŵ > w̃. That is, ŵi ↔ w̃i for all i with strict inequality somewhere. For some fixed y, let

x̂ ↓ z (ŵ, y) and x̃ ↓ z (w̃, y), which means f (x̂) ↔ y and f (x̃) ↔ y. Suppose, towards a contradiction,

that C (ŵ, y) < C (w̃, y), that is, ŵ.x̂ < w̃.x̃. Now, because ŵ > w̃, we must have ŵ.x̂ ↔ w̃.x̂. Now,

we can then say w̃.x̂ ↗ ŵ.x̂ < w̃.x̃. Which means we have found x̂ such that f (x̂) ↔ y and w̃.x̂ < w̃.x̃.

This contradicts the fact that x̃ ↓ z (w̃, y), because there is another bundle x̂ which also produces at

least y amount of output but costs strictly less than x̃. Therefore, we must have C (ŵ, y) ↔ C (w̃, y).

Which shows that C (w, y) is increasing in w.

(d) C (w, y) is homogenous of degree 1 in w.

Take arbitrary ω > 0. We have to show C (ωw, y) = ωC (w, y).

When input prices are w, the cost-minimization problem is:

min w.x
x

subject to: f (x) ↔ y

%
L %
L
Now, notice that (ωw) .x = ωwi xi = ω wi xi = ω (w.x). So when the input prices are ωw, the
i=1 i=1
cost-minimization problem can be written as:

min ω (w.x)
x

subject to: f (x) ↔ y

Now, in the two minimization problems, the constraints are the same, and satisfying that constraint,

whichever x→ minimizes w.x also minimizes ω (w.x). Therefore,

C (ωw, y) = (ωw) .x→

= ω (w.x→ )

= ωC (w, y)

↭ C (ωw, y) = ωC (w, y)

(e) C (w, y) is concave in w.

Take arbitrary ŵ, w̃ ↓ RL

++ , and ϑ ↓ [0, 1], and let wω = ϑŵ + (1 ↑ ϑ) w̃. We have to show that for any

y ↓ R+ , C (wω , y) ↔ ϑC (ŵ, y) + (1 ↑ ϑ) C (w̃, y).

Let x̂ ↓ z (ŵ, y), x̃ ↓ z (w̃, y) and xω ↓ z (wω , y). That means each of the input bundles x̂, x̃ and xω will

produce at least y amount of output. It also means when the price is ŵ, using the input bundle x̂ results

4
 in a lower cost than using any other bundle, and the same goes for w̃ and x̃. Therefore, ŵ.x̂ ↗ ŵ.xω

and w̃.x̃ ↗ w̃.xω . Now,

C (wω , y) = wω .xω

= (ϑŵ + (1 ↑ ϑ) w̃) .xω

= ϑ (ŵ.xω ) + (1 ↑ ϑ) (w̃.xω )

↔ ϑ (ŵ.x̂) + (1 ↑ ϑ) (w̃.x̃)

= ϑC (ŵ, y) + (1 ↑ ϑ) C (w̃, y)

↭ C (wω , y) ↔ ϑC (ŵ, y) + (1 ↑ ϑ) C (w̃, y)

5. Here, the firm’s daily cost minimization problem is:

min wk k + wf (Fday + Fnight )

k,Fday ,Fnight
1
subject to: (kFday ) 2 ↔ 4
1
(kFnight ) 2 ↔ 3

Now, we know that at the optimal levels, both constraints must hold with equality, otherwise the firm can

reduce Fi for whichever constraint is slack and reduce cost. So we must have:

1
(kFday ) 2 = 4
1
(kFnight ) 2
= 3

From these, we get:

kFday = 16 (1)

kFnight = 9 (2)

9 9
Dividing equation (2) by equation (1), we get = 16 , or Fnight = 16 Fday . Also, we know from equation
Fnight
Fday

5
 16
(1), we get Fday = k . That means,

9
Fday + Fnight = Fday + Fday
16
25
= Fday
16 / 
25 16
=
16 k
25
=
k

Thus, we can simplify the minimization problem to:

25
min wk k + wf
k k

Taking the FOC, we get:

/ 
1
wk + 25wf ↑ = 0
k2
25wf
wk =
k2
wf
k2 = 25
w
k
wf
↭ k↓ = 5
wk

16 16 wk
↭ Fday
↓
= =
k↓ 5 wf

9 ↓ 9 wk
↓
Fnight = F =
16 day 5 wf

Notice that the optimal capacity the firm will choose increases if fuel becomes more expensive, and decreases

if plant size becomes more expensive. Plant size and fuel are substitutes in this case.