Chapter 4 Production Theory

Advanced Microeconomic
Theory
Chapter 4: Production Theory
Outline
• Production sets and production functions
• Profit maximization and cost minimization
• Cost functions
• Aggregate supply
• Efficiency (1st and 2nd FTWE)
Advanced Microeconomic Theory 2

Production Sets and Production
Functions

Production Sets
• Let us define a production vector (or plan)
𝑦𝑦 = (𝑦𝑦1 , 𝑦𝑦2 , … , 𝑦𝑦𝐿𝐿 ) ∈ ℝ𝐿𝐿
– If, for instance, 𝑦𝑦2 > 0, then the firm is producing
positive units of good 2 (i.e., good 2 is an output).
– If, instead, 𝑦𝑦2 < 0, then the firms is producing
negative units of good 2 (i.e., good 2 is an input).
• Production plans that are technologically feasible
are represented in the production set 𝑌𝑌.
𝑌𝑌 = 𝑦𝑦 ∈ ℝ𝐿𝐿 : 𝐹𝐹(𝑦𝑦) ≤ 0
where 𝐹𝐹(𝑦𝑦) is the transformation function.
Production Sets
• 𝐹𝐹(𝑦𝑦) can also be
understood as a
production function.
• Firm uses units of 𝑦𝑦1 as
an input in order to
produce units of 𝑦𝑦2 as an
output.
• Boundary of the
production function is
any production plan 𝑦𝑦
such that 𝐹𝐹 𝑦𝑦 = 0.
– Also referred to as the
transformation frontier.
Production Sets
• For any production plan 𝑦𝑦� on the production
frontier, such that 𝐹𝐹 𝑦𝑦� = 0 , we can totally
differentiate 𝐹𝐹 𝑦𝑦� as follows
𝜕𝜕𝐹𝐹 𝑦𝑦� 𝜕𝜕𝐹𝐹 𝑦𝑦�
𝑑𝑑𝑦𝑦𝑘𝑘 + 𝑑𝑑𝑦𝑦𝑙𝑙 = 0
𝜕𝜕𝑦𝑦𝑘𝑘 𝜕𝜕𝑦𝑦𝑙𝑙
solving
�
𝜕𝜕𝐹𝐹 𝑦𝑦 �
𝜕𝜕𝐹𝐹 𝑦𝑦
𝑑𝑑𝑦𝑦𝑙𝑙 𝜕𝜕𝑦𝑦𝑘𝑘 𝜕𝜕𝑦𝑦𝑘𝑘
=− �
𝜕𝜕𝐹𝐹 𝑦𝑦 , where �
𝜕𝜕𝐹𝐹 𝑦𝑦 = 𝑀𝑀𝑀𝑀𝑀𝑀𝑙𝑙,𝑘𝑘 𝑦𝑦�
𝑑𝑑𝑦𝑦𝑘𝑘
𝜕𝜕𝑦𝑦𝑙𝑙 𝜕𝜕𝑦𝑦𝑙𝑙
– 𝑀𝑀𝑀𝑀𝑀𝑀𝑙𝑙,𝑘𝑘 𝑦𝑦� measures how much the (net) output 𝑘𝑘
can increase if the firm decreases the (net) output of
good 𝑙𝑙 by one marginal unit.
Production Sets
• What if we denote input and outputs with
different letters?
𝑞𝑞 = (𝑞𝑞1 , 𝑞𝑞2 , … , 𝑞𝑞𝑀𝑀 ) ≥ 0 outputs
𝑧𝑧 = (𝑧𝑧1 , 𝑧𝑧2 , … , 𝑧𝑧𝐿𝐿−𝑀𝑀 ) ≥ 0 inputs
where 𝐿𝐿 ≥ 𝑀𝑀.
• In this case, inputs are transformed into
outputs by the production function,
𝑓𝑓(𝑧𝑧1 , 𝑧𝑧2 , … , 𝑧𝑧𝐿𝐿−𝑀𝑀 ), i.e., 𝑓𝑓: ℝ𝐿𝐿−𝑀𝑀 → ℝ𝑀𝑀 .
Production Sets
• Example: When 𝑀𝑀 = 1 (one single output),
the production set 𝑌𝑌 can be described as
(−𝑧𝑧1 , −𝑧𝑧2 , … , −𝑧𝑧𝐿𝐿−1 , 𝑞𝑞):
𝑌𝑌 =
𝑞𝑞 ≤ 𝑓𝑓(𝑧𝑧1 , 𝑧𝑧2 , … , 𝑧𝑧𝐿𝐿−1 )
• Holding the output level fixed, 𝑑𝑑𝑑𝑑 = 0, totally
differentiate production function
𝜕𝜕𝑓𝑓 𝑧𝑧̅ 𝜕𝜕𝑓𝑓 𝑧𝑧̅
𝑑𝑑𝑧𝑧𝑘𝑘 + 𝑑𝑑𝑧𝑧𝑙𝑙 = 0
𝜕𝜕𝑧𝑧𝑘𝑘 𝜕𝜕𝑧𝑧𝑙𝑙

Production Sets
• Example (continued): and rearranging
𝜕𝜕𝑓𝑓 𝑧𝑧� 𝜕𝜕𝑓𝑓 𝑧𝑧�
𝑑𝑑𝑧𝑧𝑙𝑙 𝜕𝜕𝑧𝑧𝑘𝑘 𝜕𝜕𝑧𝑧𝑘𝑘
=− 𝜕𝜕𝑓𝑓 𝑧𝑧� , where 𝜕𝜕𝑓𝑓 𝑧𝑧� = 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 𝑧𝑧̅
𝑑𝑑𝑧𝑧𝑘𝑘
𝜕𝜕𝑧𝑧𝑙𝑙 𝜕𝜕𝑧𝑧𝑙𝑙
• 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 𝑧𝑧̅ measures the additional amount of

input 𝑘𝑘 that must be used when we marginally
decrease the amount of input 𝑙𝑙, and we want to
keep output level at 𝑞𝑞� = 𝑓𝑓 𝑧𝑧̅ .
• 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 𝑧𝑧̅ in production theory is analogous to
the 𝑀𝑀𝑅𝑅𝑆𝑆 in consumer theory, where we keep
utility constant, 𝑑𝑑𝑢𝑢 = 0.
Production Sets
• Combinations of (𝑧𝑧1 , 𝑧𝑧2 )
that produce the same
total output 𝑞𝑞 0 , i.e., z2
(𝑧𝑧1 , 𝑧𝑧2 : 𝑓𝑓(𝑧𝑧1 , 𝑧𝑧2 ) = 𝑞𝑞 0 }
is called isoquant.
• The slope of the isoquant
at (𝑧𝑧1̅ , 𝑧𝑧2̅ ) is 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 𝑧𝑧̅ .
Isoquant,
• Remember:
– 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀refers to isoquants
(and production
function).
z1
– 𝑀𝑀𝑀𝑀𝑀𝑀 refers to the
transformation function.
Production Sets
• Example: Find the𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 𝑧𝑧 for the Cobb-
𝛼𝛼 𝛽𝛽
Douglas production function 𝑓𝑓 𝑧𝑧1 , 𝑧𝑧2 = 𝑧𝑧1 𝑧𝑧2 ,
where 𝛼𝛼, 𝛽𝛽 > 0.
• The marginal product of input 1 is
𝜕𝜕𝑓𝑓 𝑧𝑧1 , 𝑧𝑧2 𝛼𝛼−1 𝛽𝛽
= 𝛼𝛼𝑧𝑧1 𝑧𝑧2
𝜕𝜕𝑧𝑧1
and that of input 2 is
𝜕𝜕𝑓𝑓 𝑧𝑧1 , 𝑧𝑧2 𝛼𝛼 𝛽𝛽−1
= 𝛽𝛽𝑧𝑧1 𝑧𝑧2
𝜕𝜕𝑧𝑧1
Production Sets
• Example (continued): Hence, the𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 𝑧𝑧 is
𝛼𝛼−1 𝛽𝛽 𝛽𝛽−(𝛽𝛽−1)
𝛼𝛼𝑧𝑧1 𝑧𝑧2 𝛼𝛼𝑧𝑧2 𝛼𝛼𝑧𝑧2
𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 𝑧𝑧 = 𝛽𝛽−1
= 𝛼𝛼−(𝛼𝛼−1) =
𝛽𝛽𝑧𝑧1𝛼𝛼 𝑧𝑧2 𝛽𝛽𝑧𝑧1 𝛽𝛽𝑧𝑧1
• For instance, for a particular vector 𝑧𝑧̅ =
1
𝑧𝑧1̅ , 𝑧𝑧2̅ = (2,3), and 𝛼𝛼 = 𝛽𝛽 = , then
2
3
𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 𝑧𝑧̅ = = 1.5
2
i.e., the slope of the isoquant evaluated at input
vector 𝑧𝑧̅ = 𝑧𝑧1̅ , 𝑧𝑧2̅ = (2,3) is −1.5.
Properties of Production Sets
1) Y is nonempty: We
have inputs and/or
outputs.
2) Y is closed: The
production set 𝑌𝑌
includes its
boundary points.

3) No free lunch: No production with no
resources.
y2
 The firm uses

amounts of input 𝑦𝑦1 Y
in order to produce y1
positive amounts of
output 𝑦𝑦2 .

2 y
3) No free lunch: violation
y2
 The firm produces y ∈ Y ∩ R+2

positive amounts of Y
good 1 and 2 (𝑦𝑦1 > 0 y1
and 𝑦𝑦2 > 0) without
the use of any inputs.

3) No free lunch: violation
y2
 The firm produces Y
positive amounts of y ∈ Y ∩ R+2

good 2 (𝑦𝑦2 > 0) with y1
zero inputs, i.e., 𝑦𝑦1 =
0.

4) Possibility of inaction
 Firm can choose to

be inactive, using no
inputs, and obtaining
no output as a result
(i.e., 0 ∈ 𝑌𝑌).

y2
 Inaction is still Set up non-

sunk cost
possible when firms Y 0 ∈Y
face fixed costs (i.e., y1
0 ∈ 𝑌𝑌).
y2
y2
 Inaction is NOT Sunk

0 ∉Y
possible when firms Y
cost
face sunk costs (i.e., y1
0 ∉ 𝑌𝑌).

5) Free disposal: if 𝑦𝑦 ∈ 𝑌𝑌 and 𝑦𝑦 ′ ≤ 𝑦𝑦, then 𝑦𝑦 ′ ∈ 𝑌𝑌.
 𝑦𝑦 ′ is less efficient than 𝑦𝑦:
– Either it produces the same amount of output with
more inputs, or less output using the same inputs.
 Then, 𝑦𝑦 ′ also belongs to the firm’s production set.
 That is, the producer can use more inputs
without the need the reduce his output:
– The producer can dispose of (or eliminate) this
additional inputs at no cost.

5) Free disposal (continued)

6) Irreversibility
y2
 Suppose that 𝑦𝑦 ∈ 𝑌𝑌
and 𝑦𝑦 ≠ 0. Then,
− 𝑦𝑦 ∉ 𝑌𝑌. y y2
y1
 “No way back” -y1
y1
Y -y2 -y ∉ Y

7) Non-increasing returns to scale: If 𝑦𝑦 ∈ 𝑌𝑌,
then 𝛼𝛼𝛼𝛼 ∈ 𝑌𝑌 for any 𝛼𝛼 ∈ 0,1 .
 That is, any feasible vector can be scaled down.

7) Non-increasing returns to scale
 The presence of fixed or sunk costs violates
non-increasing returns to scale.

8) Non-decreasing returns to scale: If 𝑦𝑦 ∈ 𝑌𝑌, then
𝛼𝛼𝛼𝛼 ∈ 𝑌𝑌 for any 𝛼𝛼 ≥ 1.
 That is, any feasible vector can be scaled up.

8) Non-decreasing returns to scale
 The presence of fixed or sunk costs do NOT
violate non-increasing returns to scale.

• Returns to scale:
– When scaling up/down
a given production plan
𝑦𝑦 = −𝑦𝑦1 , 𝑦𝑦2 :
 We connect 𝑦𝑦 with a ray
from the origin.
𝑦𝑦
 Then, the ratio 2 must
𝑦𝑦1
be maintained in all
points along the ray.
 Note that the angle of
the ray is exactly this
𝑦𝑦2
ratio .
𝑦𝑦1
9) Constant returns to scale (CRS): If 𝑦𝑦 ∈ 𝑌𝑌, then 𝛼𝛼𝛼𝛼 ∈
𝑌𝑌 for any 𝛼𝛼 ≥ 0.
α y, if α > 1 y2
 CRS is non-increasing y
and non-decreasing. α y, if α < 1
Y
y1

• Alternative graphical
representation of K
constant returns to
scale:
– Doubling 𝐾𝐾 and 𝐿𝐿 2
doubles output (i.e.,

Q=200
proportionally 1
Q=100
increase in output).
1 2 L

representation of K
increasing-returns to
scale:
increases output Q=300

more than 1
Q=200
proportionally. Q=100
1 2 L

representation of K
decreasing-returns to
scale:
increases output less
Q=200
than proportionally. 1 Q=150
Q=100
1 2 L

• Example: Let us check returns to scale in the Cobb-
𝛼𝛼 𝛽𝛽
Douglas production function 𝑓𝑓 𝑧𝑧1 , 𝑧𝑧2 = 𝑧𝑧1 𝑧𝑧2 .
Increasing all arguments by a common factor 𝜆𝜆, we
obtain
𝛽𝛽
𝑓𝑓 𝑧𝑧1 , 𝑧𝑧2 = (𝜆𝜆𝑧𝑧1 )𝛼𝛼 (𝜆𝜆𝑧𝑧2 )𝛽𝛽 = 𝜆𝜆𝛼𝛼+𝛽𝛽 𝑧𝑧1𝛼𝛼 𝑧𝑧2
– When 𝛼𝛼 + 𝛽𝛽 = 1, we have constant returns to scale;

– When 𝛼𝛼 + 𝛽𝛽 > 1, we have increasing returns to scale;
– When 𝛼𝛼 + 𝛽𝛽 < 1, we have decreasing returns to scale.

• Returns to scale in different US industries (Source:
Hsieh, 1995):
Industry 𝛼𝛼 + 𝛽𝛽
Decreasing returns Tobacco 0.51
Food 0.91
Constant returns Apparel and textile 1.01
Furniture 1.02
Electronics 1.02
Increasing returns Paper products 1.09
Petroleum and coal 1.18
Primary metal 1.24
Homogeneity of the
Returns to Scale
Production Function
𝐾𝐾 = 1 Constant Returns
𝐾𝐾 > 1 Increasing Returns
𝐾𝐾 < 1 Decreasing Returns

• The linear production function exhibits CRS as
increasing all inputs by a common factor 𝑡𝑡
yields
𝑓𝑓 𝑡𝑡𝑡𝑡, 𝑡𝑡𝑡𝑡 = 𝑎𝑎𝑎𝑎𝑎𝑎 + 𝑏𝑏𝑏𝑏𝑏𝑏 = 𝑡𝑡 𝑎𝑎𝑎𝑎 + 𝑏𝑏𝑏𝑏
≡ 𝑡𝑡𝑡𝑡(𝑘𝑘, 𝑙𝑙)
• The fixed proportion production function
𝑓𝑓 𝑘𝑘, 𝑙𝑙 = min{𝑎𝑎𝑎𝑎, 𝑏𝑏𝑏𝑏} also exhibits CRS as
𝑓𝑓 𝑡𝑡𝑡𝑡, 𝑡𝑡𝑡𝑡 = min 𝑎𝑎𝑡𝑡𝑘𝑘, 𝑏𝑏𝑏𝑏𝑏𝑏 = 𝑡𝑡 � min 𝑎𝑎𝑎𝑎, 𝑏𝑏𝑏𝑏
≡ 𝑡𝑡𝑡𝑡(𝑘𝑘, 𝑙𝑙)
• Increasing/decreasing returns to scale can be
incorporated into a production function 𝑓𝑓(𝑘𝑘, 𝑙𝑙)
exhibiting CRS by using a transformation function 𝐹𝐹(�)
𝐹𝐹 𝑘𝑘, 𝑙𝑙 = 𝑓𝑓(𝑘𝑘, 𝑙𝑙) 𝛾𝛾 , where 𝛾𝛾 > 0
• Indeed, increasing all arguments by a common factor
𝑡𝑡, yields
by CRS of 𝑓𝑓(�) 𝛾𝛾
𝐹𝐹 𝑡𝑡𝑘𝑘, 𝑡𝑡𝑙𝑙 = 𝑓𝑓(𝑡𝑡𝑘𝑘, 𝑡𝑡𝑙𝑙) 𝛾𝛾 = 𝑡𝑡 � 𝑓𝑓(𝑘𝑘, 𝑙𝑙)
= 𝑡𝑡 𝛾𝛾 � 𝑓𝑓 𝑘𝑘, 𝑙𝑙 𝛾𝛾 = 𝑡𝑡 𝛾𝛾 � 𝐹𝐹 𝑘𝑘, 𝑙𝑙
𝐹𝐹 𝑘𝑘,𝑙𝑙
• Hence,
– if 𝛾𝛾 > 1, the transformed production function
𝐹𝐹 𝑘𝑘, 𝑙𝑙 exhibits increasing returns to scale;
– if 𝛾𝛾 < 1, the transformed production function
𝐹𝐹 𝑘𝑘, 𝑙𝑙 exhibits decreasing returns to scale;

• Scale elasticity: an alternative measure of
returns to scale.
– It measures the percent increase in output due to
a 1% increase in the amounts of all inputs
𝜕𝜕𝑓𝑓(𝑡𝑡𝑡𝑡, 𝑡𝑡𝑡𝑡) 𝑡𝑡
𝜀𝜀𝑞𝑞,𝑡𝑡 = �
𝜕𝜕𝜕𝜕 𝑓𝑓(𝑘𝑘, 𝑙𝑙)
where 𝑡𝑡 denotes the common increase in all
inputs.
– Practice: Show that, if a function exhibits CRS,
then it has a scale elasticity of 𝜀𝜀𝑞𝑞,𝑡𝑡 =1.
10) Additivity (or free entry): If 𝑦𝑦 ∈ 𝑌𝑌 and 𝑦𝑦 ′ ∈
𝑌𝑌, then 𝑦𝑦 + 𝑦𝑦 ′ ∈ 𝑌𝑌.
 Interpretation: one plant produces 𝑦𝑦, while
another plant enters the market producing
𝑦𝑦 ′ . Then, the aggregate production 𝑦𝑦 +
𝑦𝑦 ′ is feasible.

11) Convexity: If 𝑦𝑦, 𝑦𝑦 ′ ∈ 𝑌𝑌 and 𝛼𝛼 ∈ 0,1 ,
then 𝛼𝛼𝛼𝛼 + (1 − 𝛼𝛼)𝑦𝑦 ′ ∈ 𝑌𝑌.
y y2
α y + (1 − α ) y ' ∈ Y
y'
Intuition: “balanced”
input-output α y + (1 − α ) y '
y'
combinations are more y1
productive than
“unbalanced” ones.

11) Convexity: violation
y
y2
α y + (1 − α ) y ' ∉ Y
Note: The convexity of αy
the production set Y
y'
maintains a close y1
relationship with the

concavity of the
production function.

11) Convexity
 With fixed costs, convexity is NOT necessarily satisfied;
 With sunk costs, convexity is satisfied.

Diminishing MRTS
• The slope of the firm’s isoquants is
𝑑𝑑𝑑𝑑 𝑓𝑓𝑙𝑙
𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑙𝑙,𝑘𝑘 = − , where 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑙𝑙,𝑘𝑘 =
𝑑𝑑𝑑𝑑 𝑓𝑓𝑘𝑘
• Differentiating 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑙𝑙,𝑘𝑘 with respect to labor

yields
𝑑𝑑𝑑𝑑 𝑑𝑑𝑑𝑑
𝜕𝜕𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑙𝑙,𝑘𝑘 𝑓𝑓𝑘𝑘 𝑓𝑓𝑙𝑙𝑙𝑙 + 𝑓𝑓𝑙𝑙𝑘𝑘 �
𝑑𝑑𝑑𝑑
− 𝑓𝑓𝑙𝑙 𝑓𝑓𝑘𝑘𝑘𝑘 + 𝑓𝑓𝑘𝑘𝑘𝑘 �
𝑑𝑑𝑑𝑑
=
𝜕𝜕𝑙𝑙 𝑓𝑓𝑘𝑘 2

Diminishing MRTS
𝑑𝑑𝑑𝑑 𝑓𝑓𝑙𝑙
• Using the fact that = − along an
𝑑𝑑𝑑𝑑 𝑓𝑓𝑘𝑘
isoquant and Young’s theorem 𝑓𝑓𝑙𝑙𝑘𝑘 = 𝑓𝑓𝑘𝑘𝑙𝑙 ,
𝑓𝑓𝑙𝑙 𝑓𝑓𝑙𝑙
𝜕𝜕𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑙𝑙,𝑘𝑘 𝑓𝑓𝑘𝑘 𝑓𝑓𝑙𝑙𝑙𝑙 − 𝑓𝑓𝑙𝑙𝑙𝑙 � 𝑓𝑓𝑘𝑘 − 𝑓𝑓𝑙𝑙 𝑓𝑓𝑘𝑘𝑘𝑘 − 𝑓𝑓𝑘𝑘𝑘𝑘 � 𝑓𝑓𝑘𝑘
=
𝜕𝜕𝜕𝜕 𝑓𝑓𝑘𝑘 2
𝑓𝑓𝑙𝑙2
𝑓𝑓𝑘𝑘 𝑓𝑓𝑙𝑙𝑙𝑙 − 𝑓𝑓𝑙𝑙𝑙𝑙 𝑓𝑓𝑙𝑙 − 𝑓𝑓𝑙𝑙 𝑓𝑓𝑘𝑘𝑘𝑘 + 𝑓𝑓𝑘𝑘𝑘𝑘 �
𝑓𝑓𝑘𝑘
=
𝑓𝑓𝑘𝑘 2

Diminishing MRTS
• Multiplying numerator and denominator by 𝑓𝑓𝑘𝑘
+ − − + + − or+
𝜕𝜕𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑙𝑙,𝑘𝑘 �2 𝑓𝑓⏞
𝑓𝑓 + 𝑓𝑓� �2
𝑓𝑓 − 2𝑓𝑓𝑙𝑙 𝑓𝑓𝑘𝑘 𝑓𝑓�
𝑘𝑘 𝑙𝑙𝑙𝑙 𝑘𝑘𝑘𝑘 𝑙𝑙 𝑙𝑙𝑙𝑙
= 3
𝜕𝜕𝜕𝜕 𝑓𝑓𝑘𝑘
• Thus,
𝜕𝜕𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑙𝑙,𝑘𝑘
– If 𝑓𝑓𝑙𝑙𝑙𝑙 > 0 (i.e., ↑ 𝑘𝑘 ⟹ ↑ 𝑀𝑀𝑃𝑃𝑙𝑙 ), then <0
𝜕𝜕𝜕𝜕
– If 𝑓𝑓𝑙𝑙𝑙𝑙 < 0, then we have
> 𝜕𝜕𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑙𝑙,𝑘𝑘 <
𝑓𝑓𝑘𝑘2 𝑓𝑓𝑙𝑙𝑙𝑙 + 𝑓𝑓𝑘𝑘𝑘𝑘 𝑓𝑓𝑙𝑙2 2𝑓𝑓𝑙𝑙 𝑓𝑓𝑘𝑘 𝑓𝑓𝑙𝑙𝑙𝑙 ⟹ 0
< 𝜕𝜕𝜕𝜕 >
Diminishing MRTS
𝑓𝑓𝑙𝑙𝑙𝑙 > 0 (↑ 𝑘𝑘 ⟹ ↑ 𝑀𝑀𝑀𝑀𝑙𝑙 ), or 𝑓𝑓𝑙𝑙𝑙𝑙 < 0 (↑ 𝑘𝑘 ⟹ ↓↓ 𝑀𝑀𝑀𝑀𝑙𝑙 )

𝑓𝑓𝑙𝑙𝑙𝑙 < 0 (↑ 𝑘𝑘 ⟹ ↓ 𝑀𝑀𝑀𝑀𝑙𝑙 ) but
small ↓ in 𝑀𝑀𝑀𝑀𝑙𝑙
Diminishing MRTS
• Example: Let us check if the production
function 𝑓𝑓 𝑘𝑘, 𝑙𝑙 = 600𝑘𝑘 2 𝑙𝑙2 − 𝑘𝑘 3 𝑙𝑙3 yields
convex isoquants.
– Marginal products:
𝑀𝑀𝑀𝑀𝑙𝑙 = 𝑓𝑓𝑙𝑙 = 1,200𝑘𝑘 2 𝑙𝑙 − 3𝑘𝑘 3 𝑙𝑙 2 > 0 iff 𝑘𝑘𝑙𝑙 < 400
𝑀𝑀𝑀𝑀𝑘𝑘 = 𝑓𝑓𝑘𝑘 = 1,200𝑘𝑘𝑙𝑙 2 − 3𝑘𝑘 2 𝑙𝑙 3 > 0 iff 𝑘𝑘𝑘𝑘 < 400
– Decreasing marginal productivity:
𝜕𝜕𝑀𝑀𝑀𝑀𝑙𝑙
= 𝑓𝑓𝑙𝑙𝑙𝑙 = 1,200𝑘𝑘 2 − 6𝑘𝑘 3 𝑙𝑙 < 0 iff 𝑘𝑘𝑘𝑘 > 200
𝜕𝜕𝑙𝑙
𝜕𝜕𝑀𝑀𝑀𝑀𝑘𝑘
= 𝑓𝑓𝑘𝑘𝑘𝑘 = 1,200𝑙𝑙 2 − 6𝑘𝑘𝑙𝑙 3 < 0 iff 𝑘𝑘𝑘𝑘 > 200
𝜕𝜕𝑘𝑘
Diminishing MRTS
• Example (continued):
– Is 200 < 𝑘𝑘𝑘𝑘 < 400 then sufficient condition for
diminishing 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀?
 No!
 We need 𝑓𝑓𝑘𝑘𝑘𝑘 > 0 too in order to guarantee diminishing
𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑙𝑙,𝑘𝑘 .
– Check the sign of 𝑓𝑓𝑙𝑙𝑙𝑙 :

𝑓𝑓𝑙𝑙𝑙𝑙 = 𝑓𝑓𝑘𝑘𝑙𝑙 = 2,400𝑘𝑘𝑘𝑘 − 9𝑘𝑘 2 𝑙𝑙 2 > 0 iff 𝑘𝑘𝑘𝑘 < 266

Diminishing MRTS
– Alternatively, we can represent the above
conditions by solving for 𝑙𝑙 in the above
inequalities:
400 𝜕𝜕𝑀𝑀𝑀𝑀𝑙𝑙 200
𝑀𝑀𝑀𝑀𝑙𝑙 > 0 iff 𝑙𝑙 < < 0 iff 𝑙𝑙 >
𝑘𝑘 𝜕𝜕𝜕𝜕 𝑘𝑘
400 𝜕𝜕𝑀𝑀𝑀𝑀𝑘𝑘 200
𝑀𝑀𝑀𝑀𝑘𝑘 > 0 iff 𝑙𝑙 < < 0 iff 𝑙𝑙 >
𝑘𝑘 𝜕𝜕𝜕𝜕 𝑘𝑘
and
266
𝑓𝑓𝑙𝑙𝑙𝑙 > 0 iff 𝑙𝑙 <
𝑘𝑘
Diminishing MRTS
200 266
– Hence, < 𝑙𝑙 < guarantees positive but
𝑘𝑘 𝑘𝑘
diminishing marginal products and, in addition, a
diminishing 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀.
200
– Figure: 𝑙𝑙 = is a curve decreasing in 𝑘𝑘, never
𝑘𝑘
266
crossing either axes. Similarly for 𝑙𝑙 = .
𝑘𝑘

Constant Returns to Scale
• If production function 𝑓𝑓(𝑘𝑘, 𝑙𝑙) exhibits CRS,
then increasing all inputs by a common factor
𝑡𝑡 yields
𝑓𝑓 𝑘𝑘, 𝑙𝑙 = 𝑡𝑡𝑡𝑡 𝑘𝑘, 𝑙𝑙
• Hence, 𝑓𝑓(𝑘𝑘, 𝑙𝑙) is homogenous of degree 1,
thus implying that its first-order derivatives
𝑓𝑓𝑘𝑘 𝑘𝑘, 𝑙𝑙 and 𝑓𝑓𝑙𝑙 𝑘𝑘, 𝑙𝑙
are homogenous of degree zero.
• Therefore,
𝜕𝜕𝜕𝜕(𝑘𝑘, 𝑙𝑙) 𝜕𝜕𝜕𝜕(𝑡𝑡𝑡𝑡, 𝑡𝑡𝑡𝑡)
𝑀𝑀𝑀𝑀𝑙𝑙 = =
𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕
= 𝑓𝑓𝑙𝑙 𝑘𝑘, 𝑙𝑙 = 𝑓𝑓𝑙𝑙 𝑡𝑡𝑡𝑡, 𝑡𝑡𝑡𝑡
1
• Setting 𝑡𝑡 = , we obtain
𝑙𝑙
1 𝑘𝑘 𝑘𝑘
𝑀𝑀𝑀𝑀𝑙𝑙 = 𝑓𝑓𝑙𝑙 𝑘𝑘, 𝑙𝑙 = 𝑓𝑓𝑙𝑙 𝑘𝑘, = 𝑓𝑓𝑙𝑙 ,1
𝑙𝑙 𝑘𝑘 𝑙𝑙
𝑘𝑘
• Hence, 𝑀𝑀𝑀𝑀𝑙𝑙 only depends on the ratio but not ,
𝑙𝑙
on the absolute levels of 𝑘𝑘 and 𝑙𝑙 that firm uses.
• A similar argument applies to 𝑀𝑀𝑀𝑀𝑘𝑘 .
𝑀𝑀𝑀𝑀𝑙𝑙
• Thus, 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 =
𝑀𝑀𝑀𝑀𝑘𝑘
only depends on the K
ratio of capital to Ray from the
labor. origin
• The slope of a firm’s Same MRTSl,k

isoquants coincides at q=4
any point along a ray q=3
from the origin. q=2
• Firm’s production
function is, hence, L
homothetic.
Elasticity of Substitution

• Elasticity of substitution (𝜎𝜎) measures the
proportionate change in the 𝑘𝑘/𝑙𝑙 ratio relative
to the proportionate change in the 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑙𝑙,𝑘𝑘
along an isoquant:
%∆(𝑘𝑘/𝑙𝑙) 𝑑𝑑(𝑘𝑘/𝑙𝑙) 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 𝜕𝜕ln(𝑘𝑘/𝑙𝑙)
𝜎𝜎 = = � =
%∆𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 𝑑𝑑𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 𝑘𝑘/𝑙𝑙 𝜕𝜕ln(𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀)
where 𝜎𝜎 > 0 since ratio 𝑘𝑘/𝑙𝑙 and 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 move
in the same direction.

• Both𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 and 𝑘𝑘/𝑙𝑙
K
will change as we
move from point 𝐴𝐴 to
point 𝐵𝐵. A
MRTSA
MRTSB
• 𝜎𝜎 is the ratio of these
B
changes. q = q0
• 𝜎𝜎 measures the (k/l)B

(k/l)A
curvature of the L
isoquant.
• If we define the elasticity of substitution
between two inputs to be proportionate
change in the ratio of the two inputs to the
proportionate change in 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀, we need to
hold:
– output constant (so we move along the same
isoquant), and
– the levels of other inputs constant (in case we
have more than two inputs). For instance, we fix
the amount of other inputs, such as land.

• High elasticity of
substitution (𝜎𝜎): K
– 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 does not

A
change substantially MRTSA
relative to 𝑘𝑘/𝑙𝑙. MRTSB

B
– Isoquant is relatively (k/l)A q = q0
flat.
(k/l)B
L

• Low elasticity of
substitution (𝜎𝜎): K
– 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 changes
substantially relative A
MRTSA
to 𝑘𝑘/𝑙𝑙. MRTSB
– Isoquant is relatively (k/l)A q = q0

B
sharply curved.
(k/l)B
L

Elasticity of Substitution:
Linear Production Function
• Suppose that the production function is
𝑞𝑞 = 𝑓𝑓 𝑘𝑘, 𝑙𝑙 = 𝑎𝑎𝑎𝑎 + 𝑏𝑏𝑏𝑏
• This production function exhibits constant
returns to scale
𝑓𝑓 𝑡𝑡𝑘𝑘, 𝑡𝑡𝑙𝑙 = 𝑎𝑎𝑎𝑎𝑎𝑎 + 𝑏𝑏𝑏𝑏𝑏𝑏 = 𝑡𝑡 𝑎𝑎𝑎𝑎 + 𝑏𝑏𝑏𝑏
= 𝑡𝑡𝑡𝑡(𝑘𝑘, 𝑙𝑙)
𝑓𝑓 𝑘𝑘,𝑙𝑙 𝑏𝑏
• Solving for 𝑘𝑘 in 𝑞𝑞, we get 𝑘𝑘 = − 𝑙𝑙.
𝑎𝑎 𝑎𝑎
– All isoquants are straight lines
– 𝑘𝑘 and 𝑙𝑙 are perfect substitutes
Linear Production Function
• 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 (slope of the
isoquant) is constant K
as 𝑘𝑘/𝑙𝑙 changes.
%∆(𝑘𝑘/𝑙𝑙) Slope=-b/a
𝜎𝜎 = =∞
%∆𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀
0
q3
• This production q1
q2
function satisfies L
homotheticity.
Fixed Proportions Production Function
𝑞𝑞 = min 𝑎𝑎𝑘𝑘, 𝑏𝑏𝑙𝑙 𝑎𝑎, 𝑏𝑏 > 0
• Capital and labor must always be used in a fixed
ratio
– No substitution between 𝑘𝑘 and 𝑙𝑙
– The firm will always operate along a ray where 𝑘𝑘/𝑙𝑙 is
constant (i.e., at the kink!).
• Because 𝑘𝑘/𝑙𝑙 is constant (𝑏𝑏/𝑎𝑎),
%∆(𝑘𝑘/𝑙𝑙)
𝜎𝜎 = =0
%∆𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀
∞
Fixed Proportions Production Function
• 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 = ∞ for 𝑙𝑙
before the kink of the K
isoquant.
• 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 = 0 for 𝑙𝑙 after
the kink. q3/a q3
• This production q2
q1
function also satisfies
homotheticity. q3/b L

Cobb-Douglas Production Function
𝑞𝑞 = 𝑓𝑓 𝑘𝑘, 𝑙𝑙 = 𝐴𝐴𝑘𝑘 𝑎𝑎 𝑙𝑙𝑏𝑏 where 𝐴𝐴, 𝑎𝑎, 𝑏𝑏 > 0
• This production function can exhibit any
returns to scale
𝑓𝑓 𝑡𝑡𝑘𝑘, 𝑡𝑡𝑙𝑙 = 𝐴𝐴(𝑡𝑡𝑡𝑡)𝑎𝑎 (𝑡𝑡𝑡𝑡)𝑏𝑏 = 𝐴𝐴𝑡𝑡 𝑎𝑎+𝑏𝑏 𝑘𝑘 𝑎𝑎 𝑙𝑙𝑏𝑏
= 𝑡𝑡 𝑎𝑎+𝑏𝑏 𝑓𝑓(𝑘𝑘, 𝑙𝑙)
– If 𝑎𝑎 + 𝑏𝑏 = 1 ⟹ constant returns to scale
– If 𝑎𝑎 + 𝑏𝑏 > 1 ⟹ increasing returns to scale
– If 𝑎𝑎 + 𝑏𝑏 < 1 ⟹ decreasing returns to scale
• The Cobb-Douglass production function is
linear in logarithms
ln 𝑞𝑞 = ln 𝐴𝐴 + 𝑎𝑎 ln 𝑘𝑘 + 𝑏𝑏 ln 𝑙𝑙
– 𝑎𝑎 is the elasticity of output with respect to 𝑘𝑘
𝜕𝜕ln(𝑞𝑞)
𝜀𝜀𝑞𝑞,𝑘𝑘 =
𝜕𝜕ln(𝑘𝑘)
– 𝑏𝑏 is the elasticity of output with respect to 𝑙𝑙
𝜕𝜕ln(𝑞𝑞)
𝜀𝜀𝑞𝑞,𝑙𝑙 =
𝜕𝜕ln(𝑙𝑙)
• The elasticity of substitution (𝜎𝜎) for the Cobb-
Douglas production function:
– First,
𝜕𝜕𝜕𝜕 𝑎𝑎 𝑙𝑙 𝑏𝑏−1
𝑀𝑀𝑀𝑀𝑙𝑙 𝜕𝜕𝑙𝑙 𝑏𝑏𝑏𝑏𝑘𝑘 𝑏𝑏 𝑘𝑘
𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 = = = = �
𝑀𝑀𝑀𝑀𝑘𝑘 𝜕𝜕𝜕𝜕 𝑎𝑎𝑎𝑎𝑘𝑘 𝑎𝑎−1 𝑙𝑙 𝑏𝑏 𝑎𝑎 𝑙𝑙
𝜕𝜕𝑘𝑘
– Hence,
𝑏𝑏 𝑘𝑘
ln(𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀) = ln + ln
𝑎𝑎 𝑙𝑙
𝑘𝑘
– Solving for ln ,
𝑙𝑙
𝑘𝑘 𝑏𝑏
ln = ln 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 − ln
𝑙𝑙 𝑎𝑎
– Therefore, the elasticity of substitution between 𝑘𝑘

and 𝑙𝑙 is
𝑘𝑘
𝑑𝑑 ln
𝑙𝑙
𝜎𝜎 = =1
𝑑𝑑 ln 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀

CES Production Function
𝑞𝑞 = 𝑓𝑓 𝑘𝑘, 𝑙𝑙 = 𝑘𝑘𝜌𝜌 + 𝑙𝑙 𝜌𝜌 𝛾𝛾/𝜌𝜌
where 𝜌𝜌 ≤ 1, 𝜌𝜌 ≠ 0, 𝛾𝛾 > 0
– 𝛾𝛾 = 1 ⟹ constant returns to scale
– 𝛾𝛾 > 1 ⟹ increasing returns to scale
– 𝛾𝛾 < 1 ⟹ decreasing returns to scale
• Alternative representation of the CES function
𝜎𝜎−1
𝜎𝜎−1 𝜎𝜎−1 𝜎𝜎
𝑓𝑓 𝑘𝑘, 𝑙𝑙 = 𝑎𝑎𝑘𝑘 𝜎𝜎
+ 𝑏𝑏𝑙𝑙 𝜎𝜎
where 𝜎𝜎 is the elasticity of substitution.
• The elasticity of substitution (𝜎𝜎) for the CES
production function:
– First,
𝛾𝛾
𝜕𝜕𝜕𝜕 𝛾𝛾 𝜌𝜌 𝜌𝜌 −1
𝑀𝑀𝑀𝑀𝑙𝑙 𝑘𝑘 + 𝑙𝑙 𝜌𝜌 𝜌𝜌𝑙𝑙 𝜌𝜌−1
𝜌𝜌
𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 = = 𝜕𝜕𝜕𝜕 = 𝛾𝛾
𝑀𝑀𝑀𝑀𝑘𝑘 𝜕𝜕𝜕𝜕 𝛾𝛾 𝜌𝜌 𝜌𝜌 −1
𝜕𝜕𝜕𝜕 𝑘𝑘 + 𝑙𝑙 𝜌𝜌 𝜌𝜌𝑘𝑘 𝜌𝜌−1
𝜌𝜌
𝜌𝜌−1 1−𝜌𝜌
𝑙𝑙 𝑘𝑘
= =
𝑘𝑘 𝑙𝑙

– Hence,
𝑘𝑘
ln(𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀) = 𝜌𝜌 − 1 ln
𝑙𝑙
𝑘𝑘
– Solving for ln ,
𝑙𝑙
𝑘𝑘 1
ln = ln 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀
𝑙𝑙 𝜌𝜌 − 1
– Therefore, the elasticity of substitution between 𝑘𝑘
and 𝑙𝑙 is
𝑘𝑘
𝑑𝑑 ln 1
𝑙𝑙
𝜎𝜎 = =
𝑑𝑑 ln 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 𝜌𝜌 − 1
• Elasticity of Substitution in German Industries
(Source: Kemfert, 1998):
Industry 𝜎𝜎
Food 0.66
Iron 0.50
Chemicals 0.37
Motor Vehicles 0.10

• The elasticity of
substitution 𝜎𝜎 K
between 𝑘𝑘 and 𝑙𝑙 is
decreasing in scale q6
(i.e., as 𝑞𝑞 increases). q5
q4
– 𝑞𝑞0 and 𝑞𝑞1 have very
high 𝜎𝜎 q3
q2
– 𝑞𝑞5 and 𝑞𝑞6 have very q0 q1
low 𝜎𝜎 L

• The elasticity of
substitution 𝜎𝜎
between 𝑘𝑘 and 𝑙𝑙 is
increasing in scale
(i.e., as 𝑞𝑞 increases).
low 𝜎𝜎
high 𝜎𝜎

Profit Maximization

Profit Maximization
• Assumptions:
– Firms are price takers: the production plans of an
individual firm do not alter price levels 𝑝𝑝 =
𝑝𝑝1 , 𝑝𝑝2 , … , 𝑝𝑝𝐿𝐿 ≫ 0.
– The production set satisfies: non-emptiness,
closedness, and free-disposal.
• Profit maximization problem (PMP):
max 𝑝𝑝 � 𝑦𝑦
𝑦𝑦
s.t. 𝑦𝑦 ∈ 𝑌𝑌, or alternatively, 𝐹𝐹(𝑦𝑦) ≤ 0
Profit Maximization
• Profit function 𝜋𝜋(𝑝𝑝) associates to every 𝑝𝑝 the
highest amount of profits (i.e., 𝜋𝜋(𝑝𝑝) is the value
function of the PMP)
𝜋𝜋 𝑝𝑝 = max 𝑝𝑝 � 𝑦𝑦: 𝑦𝑦 ∈ 𝑌𝑌
𝑦𝑦
• And the supply correspondence 𝑦𝑦(𝑝𝑝) is the
argmax of the PMP,
𝑦𝑦 𝑝𝑝 = 𝑦𝑦 ∈ 𝑌𝑌: 𝑝𝑝 � 𝑦𝑦 = 𝜋𝜋 𝑝𝑝
where positive components in the vector 𝑦𝑦 𝑝𝑝 is
output supplied by the firm to the market, while
negative components are inputs in its production
process.
Profit Maximization
• Isoprofit line:
combinations of inputs Increasing profit
and output for which y

∇F ( y )
2
the firm obtains a y(p) Supply

correspondence
given level of profits.
• Note that slope = − MRT ( y ) 1,2
y
0
𝜋𝜋 = 𝑝𝑝2 𝑦𝑦2 − 𝑝𝑝1 𝑦𝑦1
1
=y { y : F ( y ) ≤ 0}
p y −py =
2 2 π ''1 1
Solving for 𝑦𝑦2 p y −py =

2 2 π'1 1
𝜋𝜋 0 𝑝𝑝1
𝑦𝑦2 = − 𝑦𝑦1
𝑝𝑝2
� �𝑝𝑝2
intercept slope
Profit Maximization
• We can rewrite the PMP as
max 𝑝𝑝 � 𝑦𝑦
𝑦𝑦≤𝐹𝐹(𝑦𝑦)
with associated Lagrangian
𝐿𝐿 = 𝑝𝑝 � 𝑦𝑦 − 𝜆𝜆𝐹𝐹(𝑦𝑦)

Profit Maximization
– Taking FOCs with respect to every 𝑦𝑦𝑘𝑘 , we obtain
𝜕𝜕𝜕𝜕(𝑦𝑦 ∗ )
𝑝𝑝𝑘𝑘 − 𝜆𝜆 ≤0
𝜕𝜕𝑦𝑦𝑘𝑘
where 𝐹𝐹(𝑦𝑦 ∗ ) is evaluated at the optimum, i.e.,
𝐹𝐹 𝑦𝑦 ∗ = 𝐹𝐹(𝑦𝑦(𝑝𝑝)) .
𝜕𝜕𝜕𝜕(𝑦𝑦 ∗ )
– For interior solutions, 𝑝𝑝𝑘𝑘 = or in matrix
𝜆𝜆 ,
𝜕𝜕𝑦𝑦𝑘𝑘
notation
𝑝𝑝 = 𝜆𝜆𝛻𝛻𝑦𝑦 𝐹𝐹(𝑦𝑦 ∗ )
that is, the price vector and the gradient vector
are proportional.
Profit Maximization
– Solving for 𝜆𝜆, we obtain
𝑝𝑝𝑘𝑘 𝑝𝑝𝑘𝑘 𝑝𝑝𝑙𝑙
𝜆𝜆 = 𝜕𝜕𝜕𝜕(𝑦𝑦∗ ) for every good 𝑘𝑘 ⟹ 𝜕𝜕𝜕𝜕(𝑦𝑦∗ ) = 𝜕𝜕𝜕𝜕(𝑦𝑦∗ )
𝜕𝜕𝑦𝑦𝑘𝑘 𝜕𝜕𝑦𝑦𝑘𝑘 𝜕𝜕𝑦𝑦𝑙𝑙
which can also be expressed as

𝜕𝜕𝜕𝜕(𝑦𝑦∗ )
𝑝𝑝𝑘𝑘 𝜕𝜕𝑦𝑦𝑘𝑘
= 𝜕𝜕𝜕𝜕(𝑦𝑦∗ ) (= 𝑀𝑀𝑀𝑀𝑀𝑀𝑘𝑘,𝑙𝑙 (𝑦𝑦 ∗ ))
𝑝𝑝𝑙𝑙
𝜕𝜕𝑦𝑦𝑙𝑙
– Graphically, the slope of the transformation frontier

(at the profit maximization production plan 𝑦𝑦 ∗ ),
∗ 𝑝𝑝𝑘𝑘
𝑀𝑀𝑀𝑀𝑀𝑀𝑘𝑘,𝑙𝑙 (𝑦𝑦 ), coincides with the price ratio, .
𝑝𝑝𝑙𝑙
Profit Maximization
• Are there PMPs with no supply
correspondence 𝑦𝑦 𝑝𝑝 , i.e., there is no well
defined profit maximizing vector?
– Yes.
• Example: 𝑞𝑞 = 𝑓𝑓 𝑧𝑧 = 𝑧𝑧 (i.e., every unit of
input 𝑧𝑧 is transformed into a unit of output 𝑞𝑞)

Profit Maximization: Single Output
• Production function, 𝑞𝑞 = 𝑓𝑓 𝑧𝑧 , produces a single
output from a vector 𝑧𝑧 of inputs.
max 𝑝𝑝𝑝𝑝 𝑧𝑧 − 𝑤𝑤𝑤𝑤
𝑧𝑧≥0
• The first-order conditions are
𝜕𝜕𝑓𝑓(𝑦𝑦 ∗ )
𝑝𝑝 ≤ 𝑤𝑤𝑘𝑘 or 𝑝𝑝 � 𝑀𝑀𝑀𝑀𝑧𝑧𝑘𝑘 ≤ 𝑤𝑤𝑘𝑘
𝜕𝜕𝑧𝑧𝑘𝑘
• For interior solutions, the market value of the
marginal product obtained form using additional
𝜕𝜕𝜕𝜕(𝑦𝑦 ∗ )
units of this input 𝑘𝑘, 𝑝𝑝 , must coincide with
the price of this input, 𝑤𝑤𝑘𝑘 .
• Note that for any two input, this implies
𝑤𝑤𝑘𝑘
𝑝𝑝 = 𝜕𝜕𝜕𝜕(𝑦𝑦∗) for every good 𝑘𝑘
Hence,
𝜕𝜕𝜕𝜕(𝑧𝑧∗ )
𝑤𝑤𝑘𝑘 𝜕𝜕𝑧𝑧𝑘𝑘 𝑀𝑀𝑀𝑀𝑧𝑧𝑘𝑘
= 𝜕𝜕𝜕𝜕(𝑧𝑧∗ ) = (= 𝑀𝑀𝑀𝑀𝑀𝑀𝑆𝑆𝑧𝑧𝑘𝑘,𝑧𝑧𝑙𝑙 (𝑧𝑧 ∗ ))
𝑤𝑤𝑙𝑙 𝑀𝑀𝑀𝑀𝑧𝑧𝑙𝑙
𝜕𝜕𝑧𝑧𝑙𝑙
or
𝑀𝑀𝑀𝑀𝑧𝑧𝑘𝑘 𝑀𝑀𝑀𝑀𝑧𝑧𝑙𝑙
=
𝑤𝑤𝑧𝑧𝑘𝑘 𝑤𝑤𝑧𝑧𝑙𝑙
Intuition: Marginal productivity per dollar spent on
input 𝑧𝑧𝑘𝑘 is equal to that spent on input 𝑧𝑧𝑙𝑙 .
Profit Maximization
• Example : Are there PMPs with no supply correspondence
𝑦𝑦(𝑝𝑝), i.e., there is no well defined profit maximizing
vector?
– Yes.
• If the input price 𝑝𝑝𝑧𝑧 satisfies 𝑝𝑝𝑧𝑧 ≥ 𝑝𝑝, then 𝑞𝑞 = 0 and

𝜋𝜋 𝑝𝑝 = 0.
• If the input price 𝑝𝑝𝑧𝑧 satisfies 𝑝𝑝𝑧𝑧 < 𝑝𝑝, then 𝑞𝑞 = +∞ and
𝜋𝜋 𝑝𝑝 = +∞.
– In this case, the supply correspondence is not well defined,
since you can always increase input usage, thus increasing
profits.
– Exception: if input usage is constrained in the interval 0, 𝑧𝑧̅ ,
then 𝑦𝑦 𝑝𝑝 is at the corner solution 𝑦𝑦 𝑝𝑝 = 𝑧𝑧,̅ thus implying
that the PMP is well defined.
Profit Maximization
f(z)=q q=f(z) q=f(z)
f(z)=q
Increasing profit
Increasing profit
y(p)=0
z z z
If 𝑝𝑝𝑧𝑧 < 𝑝𝑝, the firm can If 𝑝𝑝𝑧𝑧 > 𝑝𝑝, the firm chooses 𝑞𝑞 =
∆𝑞𝑞 and ∆𝜋𝜋. 𝑦𝑦 𝑝𝑝 = 0 with 𝜋𝜋 𝑝𝑝 = 0.
• When are these FOCs also sufficient?
– When the production set 𝑌𝑌 is convex! Let’s see.
• Isocost line for the firm is

𝑤𝑤1 𝑧𝑧1 + 𝑤𝑤2 𝑧𝑧2 = 𝑐𝑐̅
• Solving for 𝑧𝑧2
𝑐𝑐̅ 𝑤𝑤1
𝑧𝑧2 = − 𝑧𝑧1
𝑤𝑤
�2 � 𝑤𝑤2
intercept slope

• Convex production set
– The FOCs (necessary)

𝑤𝑤1
of 𝑀𝑀𝑅𝑅𝑅𝑅𝑅𝑅 = are
𝑤𝑤2
also sufficient.

• Non-convex production set
– the FOCs are NOT
sufficient for a
combination of (𝑧𝑧1 , 𝑧𝑧2 )
that maximize profits.
– the profit-maximizing
vector (𝑧𝑧1∗ , 𝑧𝑧2∗ ) is at a
corner solution, where
the firm uses 𝑧𝑧2 alone.

• Example: Cobb-Douglas production function
• On your own:
– Solve PMP (differentiating with respect to 𝑧𝑧1 and 𝑧𝑧2 .
– Find optimal input usage 𝑧𝑧1 (𝑤𝑤, 𝑞𝑞) and 𝑧𝑧2 (𝑤𝑤, 𝑞𝑞).
• These are referred to as “conditional factor demand
correspondences”
– Plug them into the production function to obtain
the value function, i.e., the output that arises when
the firm uses its profit-maximizing input
combination.
Properties of Profit Function
• Assume that the production set 𝑌𝑌 is closed
and satisfies the free disposal property.
1) Homog(1) in prices
𝜋𝜋 𝜆𝜆𝜆𝜆 = 𝜆𝜆𝜋𝜋 𝑝𝑝
 Increasing the prices of all inputs and outputs by a
common factor 𝜆𝜆 produces a proportional
increase in the firm’s profits.
𝜋𝜋 𝑝𝑝 = 𝑝𝑝𝑝𝑝 − 𝑤𝑤1 𝑧𝑧1 − ⋯ − 𝑤𝑤𝑛𝑛 𝑧𝑧𝑛𝑛
Scaling all prices by a common factor, we obtain
𝜋𝜋 𝜆𝜆𝜆𝜆 = 𝜆𝜆𝜆𝜆𝜆𝜆 − 𝜆𝜆𝑤𝑤1 𝑧𝑧1 − ⋯ − 𝜆𝜆𝑤𝑤𝑛𝑛 𝑧𝑧𝑛𝑛
= 𝜆𝜆 𝑝𝑝𝑝𝑝 − 𝑤𝑤1 𝑧𝑧1 − ⋯ − 𝑤𝑤𝑛𝑛 𝑧𝑧𝑛𝑛 = 𝜆𝜆𝜋𝜋 𝑝𝑝
y(p’)
2) Convex in output q
prices q=f(z) y( p)
 Intuition: the firm

obtains more profits Y
from balanced input-
output combinations,
than from unbalanced y(p)
combinations.
z
Price vector Production plan Profits

𝑝𝑝 𝑦𝑦(𝑝𝑝) 𝜋𝜋 𝑝𝑝
𝑝𝑝′ 𝑦𝑦(𝑝𝑝′ ) 𝜋𝜋 𝑝𝑝′
𝑝𝑝̅ 𝑦𝑦(𝑝𝑝)̅ 𝜋𝜋 𝑝𝑝̅ = 𝛼𝛼𝛼𝛼 𝑝𝑝 + 1 − 𝛼𝛼 𝜋𝜋 𝑝𝑝′
3) If the production set 𝑌𝑌 is convex, then
𝑌𝑌 = 𝑦𝑦 ∈ ℝ𝐿𝐿 : 𝑝𝑝 � 𝑦𝑦 ≤ 𝜋𝜋 𝑝𝑝 for all 𝑝𝑝 ≫ 0
 Intuition: the production set 𝑌𝑌 can be
represented by this “dual” set.
 This dual set specifies that, for any given
prices 𝑝𝑝, all production vectors 𝑦𝑦 generate
less profits 𝑝𝑝 � 𝑦𝑦, than the optimal
production plan 𝑦𝑦(𝑝𝑝) in the profit function
𝜋𝜋 𝑝𝑝 = 𝑝𝑝 � 𝑦𝑦(𝑝𝑝).
• All production plans
𝑧𝑧, 𝑞𝑞 below the isoprofit π = p ⋅q − w⋅ z
line yield a lower profit
level: q
q=f(z)
𝑝𝑝𝑝𝑝 − 𝑤𝑤𝑤𝑤 ≤ 𝜋𝜋 𝑝𝑝 y(p)
{y ∈  2
}
: p ⋅ q − w ⋅ z ≤ π ( p)
• The isoprofit line 𝜋𝜋 𝑝𝑝 =
𝑝𝑝𝑝𝑝 − 𝑤𝑤𝑤𝑤 can be Y
expressed as
𝜋𝜋 𝑤𝑤
𝑞𝑞 = + 𝑧𝑧
𝑝𝑝 𝑝𝑝
𝑤𝑤
– If is constant ⟹ 𝜋𝜋 � z
𝑝𝑝
is convex.
– What if it is not constant?
Let’s see next. Advanced Microeconomic Theory 93
a) Input prices are a function of input usage, i.e., 𝑤𝑤 =
𝑓𝑓(𝑧𝑧), where 𝑓𝑓′(𝑧𝑧) ≠ 0. Then, either
i. 𝑓𝑓 ′ 𝑧𝑧 < 0, and the firm gets a price discount per unit of
input from suppliers when ordering large amounts of
inputs (e.g., loans)
ii. 𝑓𝑓 ′ 𝑧𝑧 > 0, and the firm has to pay more per unit of input
when ordering large amounts of inputs (e.g., scarce
qualified labor)
b) Output prices are a function of production , i.e., 𝑝𝑝 =
𝑔𝑔(𝑞𝑞), where 𝑔𝑔′(𝑞𝑞) ≠ 0. Then, either
i. 𝑔𝑔𝑔(𝑞𝑞) < 0, and the firm offers price discounts to its
customers.
ii. 𝑔𝑔𝑔(𝑞𝑞) > 0, and the firm applies price surcharges to its
customers.
• If 𝑓𝑓 ′ 𝑧𝑧 < 0, then we
have strictly convex
isoprofit curves.
• If 𝑓𝑓 ′ 𝑧𝑧 > 0, then we
have strictly concave
isoprofit curves.
• If 𝑓𝑓 ′ 𝑧𝑧 = 0, then we
have straight
isoprofit curves.
Remarks on Profit Function
• Remark 1: the profit function is a value function,
measuring firm profits only for the profit-
maximizing vector 𝑦𝑦 ∗ .
• Remark 2: the profit function can be understood
as a support function.
– Take negative of the production set 𝑌𝑌, i.e., −𝑌𝑌
– Then, the support function of −𝑌𝑌 set is
𝜇𝜇−𝑌𝑌 𝑝𝑝 = min 𝑝𝑝 � −𝑦𝑦 : 𝑦𝑦 ∈ 𝑌𝑌
𝑦𝑦
That is, take the profits resulting form all production
vectors 𝑦𝑦 ∈ 𝑌𝑌, 𝑝𝑝 � 𝑦𝑦, then take the negative of all
these profits, 𝑝𝑝 � −𝑦𝑦 , and then choose the smallest
one.
– Of course, this is the same as maximizing the
(positive) value of the profits resulting from all
production vector 𝑦𝑦 ∈ 𝑌𝑌, 𝑝𝑝 � 𝑦𝑦.
– Therefore, the profit function, 𝜋𝜋(𝑝𝑝), is the support
of the negative production set, −𝑌𝑌,
𝜋𝜋 𝑝𝑝 = 𝜇𝜇−𝑌𝑌 𝑝𝑝

– Alternatively, the
argmax of any { y : p ⋅ y ≤ π ( p)} y2
objective function q = f ( z)
y(p)
𝑦𝑦1∗ = arg max 𝑓𝑓(𝑥𝑥) y(p’)

𝑦𝑦
coincides with the y1
argmin of the negative
of this objective { y : p '⋅ y ≤ π ( p ')}
function
𝑦𝑦2∗ = arg max −𝑓𝑓(𝑥𝑥)
𝑦𝑦
where 𝑦𝑦1∗ = 𝑦𝑦2∗ .
Properties of Supply Correspondence
1) If 𝑌𝑌 is weakly convex, then 𝑦𝑦(𝑝𝑝) is a convex set
for all 𝑝𝑝.
y2
– 𝑌𝑌 has a flat surface y(p),straight

– 𝑦𝑦(𝑝𝑝) is NOT single segment of Y
valued.
y1
{y : p ⋅ y =
π ( p)}

1) (continued) If 𝑌𝑌 is strictly convex, then 𝑦𝑦(𝑝𝑝) is
single-valued (if nonempty).
y2
q = f ( z)
Unique y(p)
y1
{y : p ⋅ y =
π ( p)}

2) Hotelling’s Lemma: If 𝑦𝑦(𝑝𝑝)̅ consists of a
single point, then 𝜋𝜋(�) is differentiable at 𝑝𝑝.̅
Moreover, 𝛻𝛻𝑝𝑝 𝜋𝜋 𝑝𝑝̅ = 𝑦𝑦(𝑝𝑝).
̅
– This is an application of the duality theorem from
consumer theory.
• If 𝑦𝑦(�) is a function differentiable at 𝑝𝑝,̅ then
𝐷𝐷𝑝𝑝 𝑦𝑦(𝑝𝑝)̅ = 𝐷𝐷𝑝𝑝2 𝜋𝜋 𝑝𝑝̅ is a symmetric and positive
semidefinite matrix, with 𝐷𝐷𝑝𝑝 𝜋𝜋 𝑝𝑝̅ 𝑝𝑝̅ = 0.
 This is a direct consequence of the law of supply.
– Since 𝐷𝐷𝑝𝑝 𝜋𝜋 𝑝𝑝̅ 𝑝𝑝̅ = 0, 𝐷𝐷𝑝𝑝 𝑦𝑦(𝑝𝑝)̅ must satisfy:
 Own substitution effects (main diagonal
̅ are non-negative, i.e.,
elements in 𝐷𝐷𝑝𝑝 𝑦𝑦(𝑝𝑝))
𝜕𝜕𝑦𝑦𝑘𝑘 (𝑝𝑝)
≥ 0 for all 𝑘𝑘
𝜕𝜕𝑝𝑝𝑘𝑘
 Cross substitution effects (off diagonal elements
̅ are symmetric, i.e.,
in 𝐷𝐷𝑝𝑝 𝑦𝑦(𝑝𝑝))
𝜕𝜕𝑦𝑦𝑙𝑙 (𝑝𝑝) 𝜕𝜕𝑦𝑦𝑘𝑘 (𝑝𝑝)
= for all 𝑙𝑙 and 𝑘𝑘
𝜕𝜕𝑝𝑝𝑘𝑘 𝜕𝜕𝑝𝑝𝑙𝑙

𝜕𝜕𝑦𝑦𝑘𝑘 (𝑝𝑝)
• ≥ 0 , which
𝜕𝜕𝑝𝑝𝑘𝑘
implies that quantities
and prices move in the
same direction,
(𝑝𝑝 − 𝑝𝑝𝑝)(𝑦𝑦 − 𝑦𝑦𝑦) ≥ 0
– The law of supply holds!

• Since there is no budget constraint, there is no
wealth compensation requirement (as opposed
to Demand theory).
– This implies that there no income effects, only
substitution effects.
• Alternatively, from a revealed preference
argument, the law of supply can be expressed as
𝑝𝑝 − 𝑝𝑝′ 𝑦𝑦 − 𝑦𝑦 ′ =
𝑝𝑝𝑦𝑦 − 𝑝𝑝𝑦𝑦 ′ + 𝑝𝑝′ 𝑦𝑦′ − 𝑝𝑝′ 𝑦𝑦 ≥ 0
where 𝑦𝑦 ∈ 𝑦𝑦(𝑝𝑝) and 𝑦𝑦 ∈ 𝑦𝑦(𝑝𝑝′ ).
Cost Minimization

Cost Minimization
• We focus on the single output case, where
– 𝑧𝑧 is the input vector
– 𝑓𝑓(𝑧𝑧) is the production function
– 𝑞𝑞 are the units of the (single) output
– 𝑤𝑤 ≫ 0 is the vector of input prices
• The cost minimization problem (CMP) is
min 𝑤𝑤 � 𝑧𝑧
𝑧𝑧≥0
s. t. 𝑓𝑓(𝑧𝑧) ≥ 𝑞𝑞
Cost Minimization
• The optimal vector of input (or factor) choices
is 𝑧𝑧(𝑤𝑤, 𝑞𝑞), and is known as the conditional
factor demand correspondence.
– If single-valued, 𝑧𝑧 𝑤𝑤, 𝑞𝑞 is a function (not a
correspondence)
– Why “conditional”? Because it represents the
firm’s demand for inputs, conditional on reaching
output level 𝑞𝑞.
• The value function of this CMP 𝑐𝑐 𝑤𝑤, 𝑞𝑞 is the
cost function.
Cost Minimization
z2
Cost
minimization
w1
slope = −
w2
z ( w, q )
Isoquant f(z)=q
{z : w ⋅ z =c(w, q)} { z : w ⋅ z > c( w, q)}

z1

Cost Minimization
• Graphically,
– For a given isoquant 𝑓𝑓 𝑧𝑧 = 𝑞𝑞, choose the isocost line
associated with the lowest cost 𝑤𝑤 � 𝑧𝑧.
– The tangency point is 𝑧𝑧 𝑤𝑤, 𝑞𝑞 .
– The isocost line associated with that combination of
inputs is
𝑧𝑧: 𝑤𝑤 � 𝑧𝑧 = 𝑐𝑐 𝑤𝑤, 𝑞𝑞
where the cost function 𝑐𝑐 𝑤𝑤, 𝑞𝑞 represents the lowest
cost of producing output level 𝑞𝑞 when input prices are
𝑤𝑤.
– Other isocost lines are associated with either:
• output levels higher than 𝑞𝑞 (with costs exceeding 𝑐𝑐 𝑤𝑤, 𝑞𝑞 ),
or
• output levels lower than 𝑞𝑞 (with costs below 𝑐𝑐 𝑤𝑤, 𝑞𝑞 ).
Cost Minimization
• The Lagrangian of the CMP is
ℒ 𝑧𝑧; 𝜆𝜆 = 𝑤𝑤𝑤𝑤 + 𝜆𝜆[𝑞𝑞 − 𝑓𝑓 𝑧𝑧 ]
• Differentiating with respect to 𝑧𝑧𝑘𝑘
𝜕𝜕𝜕𝜕(𝑧𝑧 ∗ )
𝑤𝑤𝑘𝑘 − 𝜆𝜆 ≥0
(= 0 if interior solution, 𝑧𝑧𝑘𝑘∗ )
or in matrix notation
𝑤𝑤 − 𝜆𝜆𝛻𝛻𝑓𝑓(𝑧𝑧 ∗ ) ≥ 0
Cost Minimization
• From the above FOCs,
𝜕𝜕𝜕𝜕 𝑧𝑧 ∗
𝑤𝑤𝑘𝑘 𝑤𝑤𝑘𝑘 𝜕𝜕𝑧𝑧𝑘𝑘 ∗ ))
= 𝜆𝜆 ⟹ = (= 𝑀𝑀𝑅𝑅𝑅𝑅𝑅𝑅(𝑧𝑧
𝜕𝜕𝜕𝜕(𝑧𝑧 ∗ ) 𝑤𝑤𝑙𝑙 𝜕𝜕𝜕𝜕 𝑧𝑧 ∗
• Alternatively,
𝜕𝜕𝜕𝜕 𝑧𝑧 ∗ 𝜕𝜕𝜕𝜕 𝑧𝑧 ∗
=
𝑤𝑤𝑘𝑘 𝑤𝑤𝑙𝑙
at the cost-minimizing input combination, the marginal
product per dollar spent on input 𝑘𝑘 must be equal that
of input 𝑙𝑙.
Cost Minimization
• Sufficiency: If the
production set is convex,
then the FOCs are also Cost-minimizing, z(w,q)
sufficient. z2 Isoprofit line
• A non-convex production {=
z : w⋅ z cˆ, where cˆ > c( w, q)}
set:
– The input combinations
satisfying the FOCs are ẑ
NOT a cost-minimizing {z : w ⋅ z =c(w, q)}
input combination
𝑧𝑧(𝑤𝑤, 𝑞𝑞).
– The cost-minimizing
combination of inputs z1
𝑧𝑧(𝑤𝑤, 𝑞𝑞) occurs at the
corner.
Cost Minimization
• Lagrange multiplier: 𝜆𝜆 can be interpreted as
the cost increase that the firm experiences
when it needs to produce a higher level 𝑞𝑞.
– Recall that, generally, the Lagrange multiplier
represents the variation in the objective function
that we obtain if we relax the constraint (e.g.,
wealth in UMP, utility level we must reach in the
EMP).
• Therefore, 𝜆𝜆 is the marginal cost of
production: the marginal increase in the firm’s
costs form producing additional units.
Cost Minimization: SE and OE Effects
• Comparative statics of 𝑧𝑧(𝑤𝑤, 𝑞𝑞): Let us analyze
the effects of an input price change. Consider
two inputs, e.g., labor and capital. When the
price of labor, 𝑤𝑤, falls, two effects occur:
– Substitution effect: if output is held constant,
there will be a tendency for the firm to substitute
𝑙𝑙 for 𝑘𝑘.
– Output effect: a reduction in firm’s costs allows it
to produce larger amounts of output (i.e., higher
isoquant), which entails the use of more units of 𝑙𝑙
for 𝑘𝑘.
• Substitution effect:
– 𝑧𝑧 0 (𝑤𝑤, 𝑞𝑞) solves CMP at
the initial prices. K
– ↓ in wages ⟹ isocost
line pivots outwards.
z0(w,q)
– To reach 𝑞𝑞, push the 2nd step
new isocost inwards in a

parallel fashion. z1(w,q)
– 𝑧𝑧1 (𝑤𝑤, 𝑞𝑞) solves CMP at f(z)=q, isoquant
the new input prices

(for output level 𝑞𝑞). Substitution effect
st
∇w 1 step
L
– At 𝑧𝑧1 (𝑤𝑤, 𝑞𝑞), firm uses

more 𝑙𝑙 and less 𝑘𝑘.
• Substitution effect (SE):
– increase in labor K
demand from 𝐿𝐿𝐴𝐴 to 𝐿𝐿𝐵𝐵 . TC
– same output as before r
the input price change. KA A
• Output effect (OE): B

KB
ep
C
– increase in labor f(z)=q1, where q1>q0
st
rd
KC
3
demand from 𝐿𝐿𝐵𝐵 to 𝐿𝐿𝐶𝐶 . f(z)=q0, isoquant
p)
– output level increases,
e
st
nd
(2
LA LB TC LC TC
L
total cost is the same as SE
w0
OE
∇w st
(1 step) w1
before the input price TE
change.
Cost Minimization: Own-Price Effect
• We have two concepts of demand for any
input
– the conditional demand for labor, 𝑙𝑙 𝑐𝑐 (𝑟𝑟, 𝑤𝑤, 𝑞𝑞)
 𝑙𝑙𝑐𝑐 (𝑟𝑟, 𝑤𝑤, 𝑞𝑞) solves the CMP
– the unconditional demand for labor, 𝑙𝑙(𝑝𝑝, 𝑟𝑟, 𝑤𝑤)
 𝑙𝑙(𝑝𝑝, 𝑟𝑟, 𝑤𝑤) solves the PMP
• At the profit-maximizing level of output, i.e.,
𝑞𝑞(𝑝𝑝, 𝑟𝑟, 𝑤𝑤), the two must coincide
𝑙𝑙 𝑝𝑝, 𝑟𝑟, 𝑤𝑤 = 𝑙𝑙 𝑐𝑐 𝑟𝑟, 𝑤𝑤, 𝑞𝑞 = 𝑙𝑙 𝑐𝑐 (𝑟𝑟, 𝑤𝑤, 𝑞𝑞(𝑝𝑝, 𝑟𝑟, 𝑤𝑤))
• Differentiating with respect to 𝑤𝑤 yields
(+) (−)
𝜕𝜕𝜕𝜕 𝑝𝑝, 𝑟𝑟, 𝑤𝑤 𝜕𝜕𝑙𝑙 𝑐𝑐 𝑟𝑟, 𝑤𝑤, 𝑞𝑞 �
𝜕𝜕𝑙𝑙 𝑐𝑐 (𝑟𝑟, 𝑤𝑤, 𝑞𝑞) 𝜕𝜕𝜕𝜕
= + �
𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕
𝑆𝑆𝑆𝑆 (−) 𝑂𝑂𝑂𝑂 (−)
𝑇𝑇𝑇𝑇 (−)

• Since 𝑇𝑇𝑇𝑇 > 𝑆𝑆𝑆𝑆, the
unconditional labor w
demand is flatter lc(v,w,q1) lc(v,w,q2)
than the conditional

labor demand. A
• Both 𝑆𝑆𝑆𝑆 and 𝑂𝑂𝑂𝑂 are C

negative. B
– Giffen paradox from

consumer theory SE OE
TE
z
cannot arise in
production theory.
Cost Minimization: Cross-Price Effect
• No definite statement can be made about cross-
price (CP) effects.
– A fall in the wage will lead the firm to substitute away
from capital.
– The output effect will cause more capital to be
demanded as the firm expands production.
(+) (−)
𝜕𝜕𝜕𝜕 𝑝𝑝, 𝑟𝑟, 𝑤𝑤 𝜕𝜕𝑘𝑘 𝑐𝑐 𝑟𝑟, 𝑤𝑤, 𝑞𝑞 �
𝜕𝜕𝑘𝑘 𝑐𝑐 (𝑟𝑟, 𝑤𝑤, 𝑞𝑞) 𝜕𝜕𝜕𝜕
= + �
𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕
𝐶𝐶𝐶𝐶 𝑇𝑇𝑇𝑇 + or (−) 𝐶𝐶𝐶𝐶 𝑆𝑆𝑆𝑆 (+) 𝐶𝐶𝐶𝐶 𝑂𝑂𝑂𝑂 (−)
• The + cross-price
OE completely offsets w k c (r , w, q1 ) k c (r , w, q2 )
the − cross-price
SE, leading to a w
A
positive cross-price ∇w
B C
TE.
w1
k ( p, r , w)
SE
OE K
TE

• The + cross-price
OE only partially w k c (r , w, q1 )
offsets the − cross- A k c (r , w, q2 )

w
price SE, leading to a
negative cross-price ∇w
B
TE.
w1 C
k ( p, r , w)
SE
OE K
TE

Properties of Cost Function
• Assume that the production set 𝑌𝑌 is closed
and satisfies the free disposal property.
1) 𝑐𝑐(𝑤𝑤, 𝑞𝑞) is Homog(1) in 𝑤𝑤
 That is, increasing all input prices by a common
factor 𝜆𝜆 yields a proportional increase in the
minimal costs of production:
𝑐𝑐 𝜆𝜆𝑤𝑤, 𝑞𝑞 = 𝜆𝜆𝑐𝑐(𝑤𝑤, 𝑞𝑞)
since 𝑐𝑐(𝑤𝑤, 𝑞𝑞) represents the minimal cost of
producing a given output 𝑞𝑞 at input prices 𝑤𝑤.
 An increase in all
input prices (w1, w2) z2
by the same c(λ w, q) c( w, q)

=
proportion λ, λ w2
∆w2
w2
produces a parallel c( w, q)
λ w2
z ( w, q )
downward shift in the f(z)=q

firm's isocost line.
c( w, q) c(λ w, q) c( w, q)
=
z1
λ w1 λ w1 w1
∆w1

2) 𝑐𝑐(𝑤𝑤, 𝑞𝑞) is non-decreasing in 𝑞𝑞.
 Producing higher z2
output levels implies a c( w, q1 )
weakly higher minimal w2
cost of production
c( w, q0 ) z ( w, q1 )
 If 𝑞𝑞1 > 𝑞𝑞0 , then it w2
f(z)=q1
must be z ( w, q0 )
𝑐𝑐(𝑤𝑤, 𝑞𝑞1 ) > 𝑐𝑐(𝑤𝑤, 𝑞𝑞0 ) f(z)=q0
c( w, q0 ) c( w, q1 ) z1
w1 w1

3) If the set 𝑧𝑧 ≥ 0: 𝑓𝑓(𝑧𝑧) ≥ 𝑞𝑞 is convex for
every 𝑞𝑞, then the production set can be
described as
−𝑧𝑧, 𝑞𝑞 : 𝑤𝑤 � 𝑧𝑧 ≥ 𝑐𝑐 𝑤𝑤, 𝑞𝑞
𝑌𝑌 =
for every 𝑤𝑤 ≫ 0

 Take 𝑓𝑓 𝑧𝑧 = 𝑞𝑞.
 For input prices 𝑤𝑤 =
(𝑤𝑤1 , 𝑤𝑤2 ), find 𝑐𝑐(𝑤𝑤, 𝑞𝑞) by
solving CMP. z2
 For input prices 𝑤𝑤 ′ =

(𝑤𝑤1′ , 𝑤𝑤2′ ), find 𝑐𝑐(𝑤𝑤′, 𝑞𝑞) by −w
z(w,q)
solving CMP. slope = 1
w2
 The intersection of “more { z : w ⋅ z =c(w, q)}

z(w’,q)
f(z)=q0
costly” input
combinations 𝑤𝑤 � 𝑧𝑧 ≥ slope =
− w1 '
w2 '
{ z : w '⋅ z =c( w ', q)}
𝑐𝑐 𝑤𝑤, 𝑞𝑞 , for every input z1
prices 𝑤𝑤 ≫ 0, describes
set 𝑓𝑓 𝑧𝑧 ≥ 𝑞𝑞.
Properties of Conditional Factor
Demand Correspondence
1) 𝑧𝑧(𝑤𝑤, 𝑞𝑞) is Homog(0) in 𝑤𝑤.
 That is, increasing z2
input prices by the {z ≥ 0 : f ( z ) ≥ q}

z(w,q)=(z1(w,q),z2(w,q))
c( w, q )
same factor 𝜆𝜆 does w2
not alter the firm’s

c( w, q)
λ w2
demand for inputs Isocost curve Isoquant

f(z)=q
at all,
z1
𝑧𝑧 𝜆𝜆𝑤𝑤, 𝑞𝑞 = 𝑧𝑧(𝑤𝑤, 𝑞𝑞) c( w, q ) c( w, q )
λ w1 w1

2) If the set {𝑧𝑧 ≥
0: 𝑓𝑓(𝑧𝑧) ≥ 𝑞𝑞} is z2
strictly convex, then { z ≥ 0 : f ( z ) ≥ q}
the firm's demand
correspondence
Isocost curve
𝑧𝑧(𝑤𝑤, 𝑞𝑞) is single Unique Isoquant
valued. z(w,q) f(z)=q
z1

2) (continued)
If the set {𝑧𝑧 ≥ z2
0: 𝑓𝑓(𝑧𝑧) ≥ 𝑞𝑞 } is { z ≥ 0 : f ( z ) ≥ q}
weakly convex, then
the demand Isocost curve
correspondence Set of Isoquant
z(w,q) f(z)=q
𝑧𝑧(𝑤𝑤, 𝑞𝑞) is not a
single-valued, but a z1
convex set.
3) Shepard’s lemma: If 𝑧𝑧(𝑤𝑤,𝑞𝑞)
� consists of a
single point, then 𝑐𝑐(𝑤𝑤, 𝑞𝑞) is differentiable
with respect to 𝑤𝑤 at, 𝑤𝑤,
� and
𝛻𝛻𝑤𝑤 𝑐𝑐 𝑤𝑤,𝑞𝑞
� = 𝑧𝑧(𝑤𝑤,𝑞𝑞)
�

4) If 𝑧𝑧(𝑤𝑤, 𝑞𝑞) is differentiable at 𝑤𝑤,
� then
2
𝐷𝐷𝑤𝑤 𝑐𝑐 𝑤𝑤, � 𝑞𝑞 is a symmetric and
� 𝑞𝑞 = 𝐷𝐷𝑤𝑤 𝑧𝑧 𝑤𝑤,
negative semidefinite matrix, with
𝐷𝐷𝑤𝑤 𝑧𝑧 𝑤𝑤,
� 𝑞𝑞 � 𝑤𝑤� = 0.
� 𝑞𝑞 is a matrix representing how the
 𝐷𝐷𝑤𝑤 𝑧𝑧 𝑤𝑤,
firm’s demand for every unit responds to
changes in the price of such input, or in the
price of the other inputs.

4) (continued)
 Own substitution effects are non-positive,
𝜕𝜕𝑧𝑧𝑘𝑘 (𝑤𝑤,𝑞𝑞)
≤ 0 for every input 𝑘𝑘
𝜕𝜕𝑤𝑤𝑘𝑘
i.e., if the price of input 𝑘𝑘 increases, the
firm’s factor demand for this input
decreases.
 Cross substitution effects are symmetric,
𝜕𝜕𝑧𝑧𝑘𝑘 (𝑤𝑤,𝑞𝑞) 𝜕𝜕𝑧𝑧𝑙𝑙 (𝑤𝑤,𝑞𝑞)
= for all inputs 𝑘𝑘 and 𝑙𝑙
𝜕𝜕𝑤𝑤𝑙𝑙 𝜕𝜕𝑤𝑤𝑘𝑘
Properties of Production Function
1) If 𝑓𝑓(𝑧𝑧) is Homog(1) (i.e., if 𝑓𝑓(𝑧𝑧) exhibits
constant returns to scale), then 𝑐𝑐(𝑤𝑤, 𝑞𝑞) and
𝑧𝑧(𝑤𝑤, 𝑞𝑞) are Homog(1) in 𝑞𝑞.
 Intuitively, if 𝑓𝑓(𝑧𝑧) exhibits CRS, then an
increase in the output level we seek to reach
induces an increase of the same proportion
in the cost function and in the demand for
inputs. That is,
𝑐𝑐 𝑤𝑤, 𝜆𝜆𝑞𝑞 = 𝜆𝜆𝑐𝑐(𝑤𝑤, 𝑞𝑞)
and
𝑧𝑧 𝑤𝑤, 𝜆𝜆𝑞𝑞 = 𝜆𝜆𝑧𝑧(𝑤𝑤, 𝑞𝑞)
 𝜆𝜆 = 2 implies that
demand for inputs z2
doubles
𝑧𝑧 𝑤𝑤, 2𝑞𝑞 = 2𝑧𝑧(𝑤𝑤, 𝑞𝑞)
z(w,q’)=2z(w,q)
and that minimal costs λ=2
2
q'=20units
also double 1 z(w,q)

q=10units
𝑐𝑐 𝑤𝑤, 2𝑞𝑞 = 2𝑐𝑐 𝑤𝑤, 𝑞𝑞
1 2 c(w,q) z1
λ=2 c(w,q')=2c(w,q)

2) If 𝑓𝑓(𝑧𝑧) is concave, then 𝑐𝑐(𝑤𝑤, 𝑞𝑞) is convex
function of 𝑞𝑞 (i.e., marginal costs are non-
decreasing in 𝑞𝑞).
 More compactly,
𝜕𝜕 2 𝑐𝑐(𝑤𝑤, 𝑞𝑞)
2
≥0
𝜕𝜕𝑞𝑞
𝜕𝜕𝑐𝑐(𝑤𝑤,𝑞𝑞)
or, in other words, marginal costs
𝜕𝜕𝑞𝑞
are weakly increasing in 𝑞𝑞.
2) (continued)
 Firm uses more inputs z2
when raising output
from 𝑞𝑞 2 to 𝑞𝑞3 than
from 𝑞𝑞1 to 𝑞𝑞2 . z(w,q3)
z23
 Hence, q3=30units
𝑐𝑐(𝑤𝑤, 𝑞𝑞 3 ) − 𝑐𝑐(𝑤𝑤, 𝑞𝑞 2 ) > z22 z(w,q2)
𝑐𝑐(𝑤𝑤, 𝑞𝑞 2 ) − 𝑐𝑐(𝑤𝑤, 𝑞𝑞1 ) z(w,q1)
z12 q2=20units
 This reflects the q1=10units
convexity of the cost z11 z12 z13 2
c(w,q )
c(w,q3) z1
function 𝑐𝑐(𝑤𝑤, 𝑞𝑞) with c(w,q1)
respect to 𝑞𝑞.
Alternative Representation of
PMP

Alternative Representation of PMP
• Using the cost function 𝑐𝑐(𝑤𝑤, 𝑞𝑞), we write the
PMP as follows
max 𝑝𝑝𝑞𝑞 − 𝑐𝑐(𝑤𝑤, 𝑞𝑞)
𝑞𝑞≥0
This is useful if we have information about the
cost function, but we don’t about the
production function 𝑞𝑞 = 𝑓𝑓 𝑧𝑧 .

Alternative Representation of PMP
• Let us now solve this alternative PMP
max 𝑝𝑝𝑞𝑞 − 𝑐𝑐(𝑤𝑤, 𝑞𝑞)
𝑞𝑞≥0
• FOCs for 𝑞𝑞 ∗ to be profit maximizing are
𝜕𝜕𝜕𝜕(𝑤𝑤, 𝑞𝑞 ∗ )
𝑝𝑝 − ≤0
𝜕𝜕𝑞𝑞
and in interior solutions
𝜕𝜕𝜕𝜕 𝑤𝑤, 𝑞𝑞 ∗
𝑝𝑝 − =0
𝜕𝜕𝜕𝜕
• That is, at the interior∗ optimum 𝑞𝑞 ∗ , price equals
𝜕𝜕𝜕𝜕 𝑤𝑤,𝑞𝑞
marginal cost, .
𝜕𝜕𝜕𝜕
Firm’s Expansion Path
• The expansion path is the locus of cost-minimizing
tangencies. (Analogous to the wealth expansion path
in consumer theory)
K
• The curve shows how
Expansion path
inputs increase as output
increases.
• Expansion path is k c
q
positively sloped.
2
2
• Both 𝑘𝑘 and 𝑙𝑙 are normal k c

1
goods, i.e., k q
c
0 1
q0
𝜕𝜕𝑘𝑘 𝑐𝑐 (𝑤𝑤,𝑞𝑞) 𝜕𝜕𝑙𝑙 𝑐𝑐 (𝑤𝑤,𝑞𝑞)
≥ 0, ≥0 l0c l1c l2c
c(w,q1) c(w,q2)
L
𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 c(w,q0)
• If the firm’s expansion path is a straight line:
– All inputs must increase at a constant proportion as
firm increases its output.
– The firm’s production function exhibits constant
returns to scale and it is, hence, homothetic.
– If the expansion path is straight and coincides with the
45-degree line, then the firm increases all inputs by
the same proportion as output increases.
• The expansion path does not have to be a straight
line.
– The use of some inputs may increase faster than
others as output expands
• Depends on the shape of the isoquants.
• The expansion path does
not have to be upward
sloping.
– If the use of an input falls
as output expands, that
input is an inferior input.
• 𝑘𝑘 is normal
𝜕𝜕𝑘𝑘 𝑐𝑐 (𝑤𝑤, 𝑞𝑞)
≥0
𝜕𝜕𝜕𝜕
but 𝑙𝑙 is inferior (at higher
levels of output)
𝜕𝜕𝑙𝑙𝑐𝑐 (𝑤𝑤, 𝑞𝑞)
<0
𝜕𝜕𝜕𝜕 Advanced Microeconomic Theory 143
• Are there inferior inputs out there?
– We can identify inferior inputs if the list of inputs used
by the firms is relatively disaggregated.
– For instance, we can identify following categories:
CEOs, executives, managers, accountants, secretaries,
janitors, etc.
– These inputs do not increase at a constant rate as the
firm increases output (i.e., expansion path would not
be a straight line for all increases in 𝑞𝑞).
– After reaching a certain scale, the firm might buy a
powerful computer with which accounting can be
done using fewer accountants.
Cost and Supply: Single Output
• Let us assume a given vector of input prices 𝑤𝑤
�≫
0. Then, 𝑐𝑐(𝑤𝑤,
� 𝑞𝑞) can be reduced to 𝐶𝐶(𝑞𝑞). Then,
average and marginal costs are
𝐶𝐶(𝑞𝑞) 𝜕𝜕𝜕𝜕(𝑞𝑞)
𝐴𝐴𝐶𝐶 𝑞𝑞 = and 𝑀𝑀𝑀𝑀 = 𝐶𝐶′ 𝑞𝑞 =
𝑞𝑞 𝜕𝜕𝜕𝜕
• Hence, the FOCs of the PMP can be expressed as
𝑝𝑝 ≤ 𝐶𝐶𝐶 𝑞𝑞 , and in interior solutions 𝑝𝑝 = 𝐶𝐶𝐶 𝑞𝑞
i.e., all output combinations such that 𝑝𝑝 = 𝐶𝐶𝐶 𝑞𝑞
are the (optimal) supply correspondence of the
firm 𝑞𝑞 𝑝𝑝 .
• We showed that the cost function 𝑐𝑐(𝑤𝑤, 𝑞𝑞) is
homogenous of degree 1 in input prices, 𝑤𝑤.
– Can we extend this property to the AC and MC?
Yes!
– For average cost function,
𝐶𝐶(𝑡𝑡𝑡𝑡, 𝑞𝑞) 𝑡𝑡 � 𝐶𝐶(𝑤𝑤, 𝑞𝑞)
𝐴𝐴𝐴𝐴 𝑡𝑡𝑡𝑡, 𝑞𝑞 = =
𝑞𝑞 𝑞𝑞
= 𝑡𝑡 � 𝐴𝐴𝐴𝐴 𝑡𝑡𝑡𝑡, 𝑞𝑞

– For marginal cost function,
𝜕𝜕𝜕𝜕(𝑡𝑡𝑡𝑡, 𝑞𝑞) 𝑡𝑡 � 𝜕𝜕𝜕𝜕(𝑤𝑤, 𝑞𝑞)
𝑀𝑀𝑀𝑀 𝑡𝑡𝑡𝑡, 𝑞𝑞 = =
𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕
= 𝑡𝑡 � 𝑀𝑀𝑀𝑀 𝑡𝑡𝑡𝑡, 𝑞𝑞
– Isn’t this result violating Euler’s theorem? No!
 The above result states that 𝑐𝑐(𝑤𝑤, 𝑞𝑞) is homog(1) in
𝜕𝜕𝜕𝜕(𝑤𝑤,𝑞𝑞)
inputs prices, and that 𝑀𝑀𝑀𝑀 𝑤𝑤, 𝑞𝑞 = is also
𝜕𝜕𝜕𝜕
homog(1) in input prices.
 Euler’s theorem would say that: If 𝑐𝑐(𝑤𝑤, 𝑞𝑞) is
homog(1) in inputs prices, then its derivate with
𝜕𝜕𝜕𝜕(𝑤𝑤,𝑞𝑞)
respect to input prices, , must be homog(0).
𝜕𝜕𝑤𝑤
Graphical Analysis of Total Cost
Total cost
• With constant returns TC
to scale, total costs are

proportional to
output.
𝑇𝑇𝑇𝑇(𝑞𝑞) = 𝑐𝑐 � 𝑞𝑞 c
output
• Hence,
𝑇𝑇𝑇𝑇(𝑞𝑞)
𝐴𝐴𝐴𝐴(𝑞𝑞) = = 𝑐𝑐
𝑞𝑞
𝜕𝜕𝑇𝑇𝑇𝑇(𝑞𝑞)
𝑀𝑀𝐶𝐶(𝑞𝑞) = = 𝑐𝑐
𝜕𝜕𝑞𝑞
⟹ 𝐴𝐴𝐴𝐴(𝑞𝑞) = 𝑀𝑀𝑀𝑀(𝑞𝑞) Advanced Microeconomic Theory 148
• Suppose that TC starts out as concave and
then becomes convex as output increases.
– TC no longer exhibits constant returns to scale.
– One possible explanation for this is that there is a
third factor of production that is fixed as capital
and labor usage expands (e.g., entrepreneurial
skills).
– TC begins rising rapidly after diminishing returns
set in.

TC
• TC initially grows very

TC(q)
rapidly, then becomes $1,500

A
relatively flat, and for B
high production levels 0 50 q
increases rapidly again.

AC
MC(q)
• MC is the slope of the MC
$30 A’
TC curve.
AC(q)
$10 A’’
0 50 q
AC
MC MC(Q)
AC(Q)
min AC
TC becomes TC becomes q
flatter steeper
• Remark 1: AC=MC at 𝑞𝑞 = 0.
– Note that we cannot compute
𝑇𝑇𝑇𝑇 0 0
𝐴𝐴𝐴𝐴 0 = =
0 0
– We can still apply l’Hopital’s rule
𝜕𝜕𝑇𝑇𝑇𝑇(𝑞𝑞)
𝑇𝑇𝑇𝑇(𝑞𝑞) 𝜕𝜕𝑞𝑞
lim 𝐴𝐴𝐴𝐴(𝑞𝑞) = lim = lim = lim 𝑀𝑀𝐶𝐶(𝑞𝑞)
𝑞𝑞→0 𝑞𝑞→0 𝑞𝑞 𝑞𝑞→0 𝜕𝜕𝑞𝑞 𝑞𝑞→0
𝜕𝜕𝑞𝑞
– Hence, AC=MC at 𝑞𝑞 = 0, i.e., AC(0)=MC(0).

• Remark 2: When MC<AC, the AC curve
decreases, and when MC>AC, the AC curve
increases.
– Intuition: using example of grades
– If the new exam score raises your average grade, it
must be that such new grade is better than your
average grade thus far.
– If, in contrast, the new exam score lowers your
average grade, it must be that such new grade is
than your average grade thus far.
• Remark 3: AC and MC curves cross (AC=MC) at
exactly the minimum of the AC curve.
– Let us first find the minimum of the AC curve
𝑇𝑇𝐶𝐶(𝑞𝑞) 𝜕𝜕𝑇𝑇𝐶𝐶(𝑞𝑞)
𝜕𝜕𝐴𝐴𝐶𝐶(𝑞𝑞) 𝜕𝜕 𝑞𝑞 − 𝑇𝑇𝑇𝑇(𝑞𝑞) � 1
𝑞𝑞 𝜕𝜕𝜕𝜕
= =
𝜕𝜕𝑞𝑞 𝜕𝜕𝑞𝑞 𝑞𝑞 2
𝑞𝑞 � 𝑀𝑀𝑀𝑀(𝑞𝑞) − 𝑇𝑇𝑇𝑇(𝑞𝑞)
= 2
=0
𝑞𝑞
– The output that minimizes AC must satisfy
𝑇𝑇𝑇𝑇 𝑞𝑞
𝑞𝑞 � 𝑀𝑀𝑀𝑀 𝑞𝑞 − 𝑇𝑇𝑇𝑇 𝑞𝑞 = 0 ⟹ 𝑀𝑀𝑀𝑀 𝑞𝑞 = ← 𝐴𝐴𝐴𝐴 𝑞𝑞
𝑞𝑞
– Hence, 𝑀𝑀𝑀𝑀 = 𝐴𝐴𝐴𝐴 at the minimum of 𝐴𝐴𝐶𝐶.
• Decreasing returns to scale:
– an increase in the use of inputs produces a less-than-
proportional increase in output.
 production set is strictly convex
 TC function is convex
 MC and AC are increasing
q p
C(q)
slope = C '(qˆ )
C’(q)
Heavy trace
is supply
Y slope = AC (qˆ )
locus q(p)
q̂
AC(q)
(a) z -z (b) q̂ q (c) q

• Constant returns to scale:
– an increase in input usage produces a proportional
increase in output.
 production set is weakly convex
 linear TC function
 constant AC and MC functions
q C(q) p
q(p)
Y
AC(q) = C’(q)
No sales for p < MC(q)
(a) -z (b) q (c) q

• Increasing returns to scale:
– an increase in input usage can lead to a more-than-
proportional increase in output.
 production set is non-convex
 TC curve first increases, then becomes almost flat, and then
increases rapidly again as output is increased further.
C(q)
q p C’(q)
AC(q)
q(p)
(a) -z (b) q (c) q

• Let us analyze the presence of non-convexities
in the production set 𝑌𝑌 arising from:
– Fixed set-up costs, 𝐾𝐾, that are non-sunk
𝐶𝐶 𝑞𝑞 = 𝐾𝐾 + 𝐶𝐶𝑣𝑣 𝑞𝑞
where 𝐶𝐶𝑣𝑣 (𝑞𝑞) denotes variable costs
• with strictly convex variable costs
• with linear variable costs
– Fixed set-up costs that are sunk
• Cost function is convex, and hence FOCs are sufficient
• CRS technology and fixed (non-sunk) costs:
– If 𝑞𝑞 = 0, then 𝐶𝐶 𝑞𝑞 = 0, i.e., firm can recover 𝐾𝐾 if it
shuts down its operation.
– MC is constant: 𝑀𝑀𝑀𝑀 = 𝐶𝐶′ 𝑞𝑞 = 𝐶𝐶𝑣𝑣′ 𝑞𝑞 = 𝑐𝑐
𝐶𝐶(𝑞𝑞) 𝐾𝐾 𝐶𝐶𝑣𝑣 𝑞𝑞 𝐾𝐾
– AC lies above MC: 𝐴𝐴𝐴𝐴 𝑞𝑞 = = + = + 𝑐𝑐
𝑞𝑞 𝑞𝑞 𝑞𝑞 𝑞𝑞

• DRS technology and fixed (non-sunk) costs:
– MC is positive and increasing in 𝑞𝑞, and hence the slope of the TC
curve increases in 𝑞𝑞.
– in the decreasing portion of the AC curve, FC is spread over
larger 𝑞𝑞.
– in the increasing portion of the AC curve, larger average VC
offsets the lower average FC and, hence, total average cost
increases.
q p
C(q)
Cv(q)=C’(q)
AC(q)
Y AC (q )
q(q)
(a) -z (b) q q (c) q q

• DRS technology and sunk costs:
– TC curve originates at 𝐾𝐾, given that the firm must
incur fixed sunk cost 𝐾𝐾 even if it chooses 𝑞𝑞 = 0.
– supply locus considers the entire MC curve and not
only 𝑞𝑞 for which MC>AC.
q p
C(q)
C’(q)
q(p)
AC(q)
(a) -z (b) q (c) q

Short-Run Total Cost
• In the short run, the firm generally incurs
higher costs than in the long run
– The firm does not have the flexibility of input
choice (fixed inputs).
– To vary its output in the short-run, the firm must
use non-optimal input combinations
– The 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 will not be equal to the ratio of input
prices.

Short-Run vs Long-Run Total Cost
• In the short-run
�
– capital is fixed at 𝐾𝐾 K
– the firm cannot

equate 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 with
C(w,Q0)
the ratio of input r
prices.
• In the long-run A
F
– Firm can choose
input vector 𝐴𝐴, which C(w,Q0)
w
L
is a cost-minimizing
input combination.
• 𝑞𝑞 = 1 million units
– Firm chooses (𝑘𝑘1 , 𝑙𝑙1 )
K
both in the long run
and in the short run Expansion path
when 𝑘𝑘 = 𝑘𝑘1 .
• 𝑞𝑞 = 2 million units
– Short-run (point B): K2 C
B
 𝑘𝑘 = 𝑘𝑘1 does not allow K1 A
Q=2
the firm to minimize
costs.
Q=1
– Long-run (point C): 0 L1 L2 L2 L
 firm can choose cost-
minimizing input
combination.
• The difference STC(Q) when k=k1
between long-run, TC TC(q)
𝑇𝑇𝑇𝑇(𝑞𝑞), and short- B
run, 𝑆𝑆𝑆𝑆𝑆𝑆(𝑞𝑞), total

A C
costs when capital is
fixed at 𝑘𝑘 = 𝑘𝑘1 . rK1
0 1 million 2 million Q

• The long-run total
cost curve 𝑇𝑇𝑇𝑇(𝑞𝑞) can TC
STC(q) where k=k2
be derived by varying STC(q) where k=k0

TC(q)
the level of 𝑘𝑘.

STC(q) where k=k1
• Short-run total cost

curve 𝑆𝑆𝑆𝑆𝑆𝑆(𝑞𝑞) lies
above long-run total
q0 q1 q2 q
cost 𝑇𝑇𝑇𝑇(𝑞𝑞).

• Summary:
– In the long run, the firm can modify the values of
all inputs.
– In the short run, in contrast, the firm can only
modify some inputs (e.g., labor, but not capital).

• Example: Short- and long-run curves
– In the long run,
𝐶𝐶 𝑞𝑞 = 𝑤𝑤
�1 𝑧𝑧1 + 𝑤𝑤
�2 𝑧𝑧2
where both input 1 and 2 are variable.
– In the short run, input 2 is fixed at 𝑧𝑧2̅ , and thus
𝐶𝐶 𝑞𝑞|𝑧𝑧2̅ = 𝑤𝑤
�1 𝑧𝑧1 + 𝑤𝑤
�2 𝑧𝑧2̅
• This implies that the only input that the firm can modify
is input 1.
• The firm chooses 𝑧𝑧1 such that production reaches
output level 𝑞𝑞, i.e., 𝑓𝑓(𝑧𝑧1 , 𝑧𝑧2̅ ) = 𝑞𝑞.
– When the demand for input 2 is at its long-run
value, i.e., 𝑧𝑧2 (𝑤𝑤, 𝑞𝑞), then
𝐶𝐶 𝑞𝑞 = 𝐶𝐶(𝑞𝑞|𝑧𝑧2 (𝑤𝑤, 𝑞𝑞)) for every 𝑞𝑞
and also
𝐶𝐶′ 𝑞𝑞 = 𝐶𝐶′(𝑞𝑞|𝑧𝑧2 (𝑤𝑤, 𝑞𝑞)) for every 𝑞𝑞
i.e., values and slopes of long- and short-run cost
functions coincide.
– Long- and short-run curves are tangent at
𝑧𝑧2 (𝑤𝑤, 𝑞𝑞). Advanced Microeconomic Theory 169
– Since
𝐶𝐶(𝑞𝑞) ≤ 𝐶𝐶(𝑞𝑞|𝑧𝑧2 ) for any given 𝑧𝑧2 ,
then the long-run cost curve 𝐶𝐶 𝑞𝑞 is the lower
envelope of the short-run cost curves, 𝐶𝐶(𝑞𝑞|𝑧𝑧2 ).

Aggregation in Production

• Let us analyze under which conditions the
“law of supply” holds at the aggregate level.
• An aggregate production function maps
aggregate inputs into aggregate outputs
– In other words, it describes the maximum level of
output that can be obtained if the inputs are
efficiently used in the production process.

• Consider 𝐽𝐽 firms, with production sets 𝑌𝑌1 , 𝑌𝑌2 , … , 𝑌𝑌𝐽𝐽 .
• Each 𝑌𝑌𝑗𝑗 is non-empty, closed, and satisfies the free
disposal property.
• Assume also that every supply correspondence 𝑦𝑦𝑗𝑗 (𝑝𝑝) is
single valued, and differentiable in prices, 𝑝𝑝 ≫ 0.
• Define the aggregate supply correspondence as the
sum of the individual supply correspondences
𝐽𝐽 𝐽𝐽
𝑦𝑦 𝑝𝑝 = � 𝑦𝑦𝑗𝑗 𝑝𝑝 = 𝑦𝑦 ∈ ℝ𝐿𝐿 : 𝑦𝑦 = � 𝑦𝑦𝑗𝑗 𝑝𝑝
𝑗𝑗=1 𝑗𝑗=1
where 𝑦𝑦𝑗𝑗 ∈ 𝑦𝑦𝑗𝑗 (𝑝𝑝) for 𝑗𝑗 = 1,2, … , 𝐽𝐽.

• The law of supply is satisfied at the aggregate
level.
• Two ways to check it:
1) Using the derivative of every firm’s supply
correspondence with respect to prices, 𝐷𝐷𝑝𝑝 𝑦𝑦𝑗𝑗 𝑝𝑝 .
– 𝐷𝐷𝑝𝑝 𝑦𝑦𝑗𝑗 𝑝𝑝 is a symmetric positive semidefinite
matrix, for every firm 𝑗𝑗.
– Since this property is preserved under
addition, then 𝐷𝐷𝑝𝑝 𝑦𝑦 𝑝𝑝 must also define a
symmetric positive semidefinite matrix.
2) Using a revealed preference argument.
– For every firm 𝑗𝑗,
𝑝𝑝 − 𝑝𝑝𝑝 � 𝑦𝑦𝑗𝑗 𝑝𝑝 − 𝑦𝑦𝑗𝑗 𝑝𝑝′ ≥0
– Adding over 𝑗𝑗,

𝑝𝑝 − 𝑝𝑝𝑝 � 𝑦𝑦 𝑝𝑝 − 𝑦𝑦 𝑝𝑝𝑝 ≥0
– This implies that market prices and aggregate
supply move in the same direction
 the law of supply holds at the aggregate level!

• Is there a “representative producer”?
– Let 𝑌𝑌 be the aggregate production set,
𝐽𝐽
𝑌𝑌 = 𝑌𝑌1 + 𝑌𝑌2 +. . . +𝑌𝑌𝑗𝑗 = 𝑦𝑦 ∈ ℝ𝐿𝐿 : 𝑦𝑦 = � 𝑦𝑦𝑗𝑗
𝑗𝑗=1
for some 𝑦𝑦𝑗𝑗 ∈ 𝑌𝑌𝑗𝑗 and 𝑗𝑗 = 1,2, … , 𝐽𝐽.
𝐽𝐽
– Note that 𝑦𝑦 = ∑𝑗𝑗=1 𝑦𝑦𝑗𝑗 , where every 𝑦𝑦𝑗𝑗 is just a
feasible production plan of firm 𝑗𝑗, but not necessarily
firm 𝑗𝑗’s supply correspondence 𝑦𝑦𝑗𝑗 (𝑝𝑝).
– Let 𝜋𝜋 ∗ (𝑝𝑝) be the profit function for the aggregate
production set 𝑌𝑌.
– Let 𝑦𝑦 ∗ (𝑝𝑝) be the supply correspondence for the
aggregate production set 𝑌𝑌.
– Then, there exists a representative producer:
• Producing an aggregate supply 𝑦𝑦 ∗ (𝑝𝑝) that exactly coincides
𝐽𝐽
with the sum ∑𝑗𝑗=1 𝑦𝑦𝑗𝑗 𝑝𝑝 ; and
• Obtaining aggregate profits 𝜋𝜋 ∗ (𝑝𝑝) that exactly coincide with
𝐽𝐽
the sum ∑𝑗𝑗=1 𝜋𝜋𝑗𝑗 (𝑝𝑝).
– Intuition: The aggregate profit obtained by each firm

maximizing its profits separately (taking prices as
given) is the same as that which would be obtained if
all firms were to coordinate their actions (i.e., 𝑦𝑦𝑗𝑗 ’s) in
a joint PMP.
– It is a “decentralization” result: to find the solution
of the joint PMP for given prices 𝑝𝑝, it is enough to
“let each individual firm maximize its own profits”
and add the solutions of their individual PMPs.
– Key: price taking assumption
• This result does not hold if firms have market power.
• Example: oligopoly markets where firms compete in
quantities (a la Cournot).

• Firm 1 chooses 𝑦𝑦1 given 𝑝𝑝
and 𝑌𝑌1 .
• Firm 2 chooses 𝑦𝑦2 given
𝑝𝑝 and 𝑌𝑌2 .
• Jointly, the two firms
would be selecting 𝑦𝑦1 +
𝑦𝑦2 .
• The aggregate supply
correspondence 𝑦𝑦1 + 𝑦𝑦2
coincides with the supply
correspondence that a
single firm would select
given 𝑝𝑝 and 𝑌𝑌 = 𝑦𝑦1 + 𝑦𝑦2 .
Efficient Production

• Efficient production vector: a production
vector 𝑦𝑦 ∈ 𝑌𝑌 is efficient if there is no other
𝑦𝑦′ ∈ 𝑌𝑌 such that 𝑦𝑦𝑦 ≥ 𝑦𝑦 and 𝑦𝑦𝑦 ≠ 𝑦𝑦.
– That is, 𝑦𝑦 is efficient if there is no other feasible
production vector 𝑦𝑦𝑦 producing more output with
the same amount of inputs, or alternatively,
producing the same output with fewer inputs.
𝑦𝑦 is efficient ⇒ 𝑦𝑦 is on the boundary of 𝑌𝑌
𝑦𝑦 is efficient ⇍ 𝑦𝑦 is on the boundary of 𝑌𝑌

• 𝑦𝑦𝑦𝑦 produces the same
output as 𝑦𝑦, but uses y
more inputs. y
• 𝑦𝑦𝑦 uses the same
inputs as 𝑦𝑦, but y''
produces less output. y'
• 𝑦𝑦𝑦𝑦 and 𝑦𝑦𝑦 are Y
inefficient. z
• 𝑦𝑦 is efficient ⇒ 𝑦𝑦 lies
on the frontier of the
production set 𝑌𝑌.
• 𝑦𝑦 is efficient
• 𝑦𝑦𝑦 is inefficient y
– it produces the same y' y

output as 𝑦𝑦, but uses
more inputs.
Y
• Hence, 𝑦𝑦𝑦 lies on the
frontier of the z
production set 𝑌𝑌 ⇏ 𝑦𝑦𝑦 is
efficient.

Efficient Production: 1st FTWE
• 1st Fundamental Theorem of Welfare Economics
(FTWE): If 𝑦𝑦 ∈ 𝑌𝑌 is profit-maximizing for some
price vector 𝑝𝑝 ≫ 0, then 𝑦𝑦 must be efficient.
Proof: Let us prove the 1st FTWE by contradiction.
Suppose that 𝑦𝑦 ∈ 𝑌𝑌 is profit-maximizing
𝑝𝑝 � 𝑦𝑦 ≥ 𝑝𝑝 � 𝑦𝑦′ for all 𝑦𝑦𝑦 ∈ 𝑌𝑌
but 𝑦𝑦 is not efficient. That is, there is a 𝑦𝑦𝑦 ∈ 𝑌𝑌 such that
𝑦𝑦𝑦 ≥ 𝑦𝑦. If we multiply both sides of 𝑦𝑦𝑦 ≥ 𝑦𝑦 by 𝑝𝑝, we
obtain
𝑝𝑝 � 𝑦𝑦′ ≥ 𝑝𝑝 � 𝑦𝑦, since 𝑝𝑝 ≫ 0
But then, 𝑦𝑦 cannot be profit-maximizing. Contradiction!
• For the result in 1st FTWE, we do NOT need the
production set 𝑌𝑌 to be convex.
– 𝑦𝑦 is profit-maximizing ⇒ 𝑦𝑦 lies on a tangency point
p p
y y y
y
y'
Y Isoprofit line
z z
convex production set non-convex production set

• Note: the assumption 𝑝𝑝 ≫ 0 cannot be relaxed to 𝑝𝑝 ≥ 0.
– Take a production set 𝑌𝑌 with an upper flat surface.
– Any production plan in
the flat segment of 𝑌𝑌 can p y
be profit-maximizing if y
prices are 𝑝𝑝 = (0,1).
– But only 𝑦𝑦 is efficient.
– Other profit-maximizing Y
Profit
production plans to the maximizing
left of 𝑦𝑦 are NOT efficient. Production z
plans
– Hence, in order to apply
1st FTWE we need 𝑝𝑝 ≫ 0.
Efficient Production: 2nd FTWE
• The 2nd FTWE states the converse of the 1st
FTWE:
– If a production plan 𝑦𝑦 is efficient, then it must be
profit-maximizing.
• Note that, while it is true for convex
production sets, it cannot be true if 𝑌𝑌 is non-
convex.

• The 2nd FTWE is restricted to convex production sets.
• For non-convex production set: If 𝑦𝑦 is efficient ⇏
𝑦𝑦 is profit-maximizing
p p
y y y
y
y'
Y Isoprofit line
z z
convex production set non-convex production set

• 2nd FTWE: If production set 𝑌𝑌 is convex, then
every efficient production plan 𝑦𝑦 ∈ 𝑌𝑌 is profit-
maximizing production plan, for some non-
zero price vector 𝑝𝑝 ≥ 0.
Proof:
1) Take an efficient production plan, such as 𝑦𝑦 on
the boundary of 𝑌𝑌. Define the set of production
plans that are strictly more efficient than 𝑦𝑦
𝑃𝑃𝑦𝑦 = 𝑦𝑦 ′ ∈ ℝ𝐿𝐿 : 𝑦𝑦 ′ ≫ 𝑦𝑦
2) Note that 𝑌𝑌 ∩ 𝑃𝑃𝑦𝑦 ≠ ∅ and 𝑃𝑃𝑦𝑦 is convex set.
Proof (continued):
3) From the Separating
Hyperplane Theorem, y
there is some 𝑝𝑝 ≠ 0 Py
such that 𝑝𝑝 � 𝑦𝑦′ ≥ 𝑝𝑝 � 𝑦𝑦′′, y'
for 𝑦𝑦𝑦 ∈ 𝑃𝑃𝑦𝑦 and 𝑦𝑦′′ ∈ 𝑌𝑌. y
p
4) Since 𝑦𝑦𝑦 can be made
arbitrarily close to 𝑦𝑦, we Y Isoprofit line
can have 𝑝𝑝 � 𝑦𝑦 ≥ 𝑝𝑝 � 𝑦𝑦′′ z

for 𝑦𝑦𝑦𝑦 ∈ 𝑌𝑌.
5) Hence, the efficient
production plan 𝑦𝑦 must
be profit-maximizing.
• Note: we are not imposing 𝑝𝑝 ≫ 0, but 𝑝𝑝 ≥ 0.
– We just assume that
the price vector is not
y
zero for every p
component, i.e., 𝑝𝑝 ≠ y' y
(0,0, … , 0).
– Hence, the slope of
Y
the isoprofit line can Profit
be zero. maximizing
Production z
– Both 𝑦𝑦 and 𝑦𝑦′ are plans
profit-maximizing,
but only 𝑦𝑦 is efficient.
• Note: the 2nd FTWE does not allow for input
prices to be negative.
– Consider the case in which the price of input 𝑙𝑙 is
negative, 𝑝𝑝𝑙𝑙 < 0.
– Then, we would have 𝑝𝑝 � 𝑦𝑦′ < 𝑝𝑝 � 𝑦𝑦 for some
production plan 𝑦𝑦𝑦 that is more efficient than 𝑦𝑦,
i.e., 𝑦𝑦𝑦 ≫ 𝑦𝑦, with 𝑦𝑦𝑙𝑙′ − 𝑦𝑦𝑙𝑙 being sufficiently large.
– This implies that the firm is essentially “paid” for
using further amounts of input 𝑙𝑙.
– For this reason, we assume 𝑝𝑝 ≥ 0.

Chapter 4 Production Theory

Uploaded by

Document Informationclick to expand document informationnone

Document Informationclick to expand document information

Copyright:

Available Formats

Chapter 4 Production Theory

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Chapter 4 Production Theory

Uploaded by

Copyright:

Available Formats

Advanced Microeconomic

Advanced Microeconomic Theory 2

Advanced Microeconomic Theory 3

Advanced Microeconomic Theory 8

• 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 𝑧𝑧̅ measures the additional amount of

Advanced Microeconomic Theory 13

 The firm uses

Advanced Microeconomic Theory 14

 The firm produces y ∈ Y ∩ R+2

Advanced Microeconomic Theory 15

 The firm produces Y

positive amounts of y ∈ Y ∩ R+2

Advanced Microeconomic Theory 16

 Firm can choose to

Advanced Microeconomic Theory 17

 Inaction is still Set up non-

face fixed costs (i.e., y1

 Inaction is NOT Sunk

face sunk costs (i.e., y1

Advanced Microeconomic Theory 19

Advanced Microeconomic Theory 20

Advanced Microeconomic Theory 21

Advanced Microeconomic Theory 22

Advanced Microeconomic Theory 23

Advanced Microeconomic Theory 24

Advanced Microeconomic Theory 25

Advanced Microeconomic Theory 26

Advanced Microeconomic Theory 28

doubles output (i.e.,

Advanced Microeconomic Theory 29

increases output Q=300

Advanced Microeconomic Theory 30

Advanced Microeconomic Theory 31

– When 𝛼𝛼 + 𝛽𝛽 = 1, we have constant returns to scale;

Advanced Microeconomic Theory 32

Advanced Microeconomic Theory 34

Advanced Microeconomic Theory 37

Advanced Microeconomic Theory 39

combinations are more y1

Advanced Microeconomic Theory 40

relationship with the

Advanced Microeconomic Theory 41

Advanced Microeconomic Theory 42

• Differentiating 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑙𝑙,𝑘𝑘 with respect to labor

Advanced Microeconomic Theory 43

Advanced Microeconomic Theory 44

𝑓𝑓𝑙𝑙𝑙𝑙 > 0 (↑ 𝑘𝑘 ⟹ ↑ 𝑀𝑀𝑀𝑀𝑙𝑙 ), or 𝑓𝑓𝑙𝑙𝑙𝑙 < 0 (↑ 𝑘𝑘 ⟹ ↓↓ 𝑀𝑀𝑀𝑀𝑙𝑙 )

– Check the sign of 𝑓𝑓𝑙𝑙𝑙𝑙 :

Advanced Microeconomic Theory 48

Advanced Microeconomic Theory 50

• The slope of a firm’s Same MRTSl,k

Advanced Microeconomic Theory 54

Advanced Microeconomic Theory 55

• 𝜎𝜎 measures the (k/l)B

Advanced Microeconomic Theory 57

– 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 does not

relative to 𝑘𝑘/𝑙𝑙. MRTSB