Advanced Microeconomics

Yu Gao
17, October 2024
Table of contents

1. Theory of firms
Technology: production set and production function
Profit maximization problem (PMP)
Profit functions
Cost minimization
Cost function

2. The geometry of cost

3. Aggregation in production

1
Theory of firms
Theory of consumers and theory of firms

Firms are special “consumers”. They purchase “inputs”. Their “utility”

comes from profits.

Consumers Firms
Utility function Production function
Indifference curve Iso-quant curve
Utility maximizing Profit maximizing
Expenditure minimizing Cost minimizing

2
Technology: production set and
production function
Production function

Let us define a production vector (or plan)

y = (y1 , y2 , ..., yL ) ∈ RL

• Suppose that L=5. Then y = (−5, 2, −6, 3, 0) means that 2 and 3

units of goods 2 and 4, respectively, are produced, which 5 and 6
units of goods 1 and 3, respectively, are used. Good 5 is neither
produced nor used as an input in this production vector.

Production plans that are technologically feasible are represented in the

production set Y.
Y = {y ∈ RL : F(y) ≤ 0}
where F(y) is the transformation function.

3
Production function

Take the example of two goods: L=2.

• In the left-hand segment of
the figure, y1 < 0 and y2 > 0.
Therefore, y1 is “input”
whereas y2 is “output”.
• In the right-hand segment of
the figure, the firm essentially
“reverts” its production
process in order to use units
of y2 to produce units of y1 .

4
Production function

The set of boundary points of Y, {y ∈ RL : F(y) = 0}, is known as the

transformation frontier.

• We can totally differentiate along F(y)=0, as follows:

∂F(ȳ) ∂F(ȳ)
dF(ȳ) = dyk + dyl = 0
∂yk ∂yl
dyl ∂F(ȳ)/∂yk
=−
dyk ∂F(ȳ)/∂yl

where the term on the right hand side is

∂F(ȳ)/∂yk
≡ MRTl,k (ȳ)
∂F(ȳ)/∂yl

• MRTl,k is the marginal rate of transformation between good l and

k.

5
Technologies with distinct inputs and outputs

In many actual production processes, the set of goods that can be

outputs is distinct from the set that can be inputs.
For simplicity, assuming the firms has only one output, then

q = f(z1 , z2 , ..., zn ) = f(z)

where z is a vector of inputs, for example workers, computers, machines,

land, etc. f(z) is the production function.

• Holding the output level fixed, dq = 0, totally differentiate

production function (assuming there are only capital and labor)

∂f(z̄) ∂f(z̄)
dzk + dzl = 0
∂zk ∂zl

6
Technologies with distinct inputs and outputs

• Rearranging,
∂f(z̄)
dzl ∂z
= − ∂f(z̄k)
dzk
∂zl

where
∂f(z̄)
∂zk
∂f(z̄)
≡ MRTS(z̄)
∂zl

• MRTS(z̄) measures the additional amount of input k that must be

used when we marginally decrease the amount of input l, and we
want to keep output level q̄ = f(z̄).
• MRTS(z̄) in production theory is analogous to the MRS in consumer
theory, where we keep utility constant, du = 0.

7
Iso-quant curves

In consumer theories, given a utility function, we can draw a series of

indifference curves.
Likewise, given a production function, we can draw a series of iso-quant
curves (for simplicity, assuming two inputs: K and L):

8
Marginal rate of technical substitution (MRTS)

In the two-dimensional case, we can increase the amount of input 1 and

decrease the amount of input 2 so as to maintain a constant level of
output.

Note:
• MRTS refers to iso-quant curves (and production function).
• MRT refers to the transformation function and transformation
frontier. 9
Returns to scale
Returns to scale

What will happen to the level of output if we scale all inputs up or down
by some amount t ≥ 0?
Constant returns to scale (CRS) :
• f(tz) = tf(z) for all t ≥ 0; i.e., the production function f(z) is
homogeneous of degree 1.

10
Returns to scale

Increasing returns to scale (IRS) :

• f(tz) > tf(z) for all t > 1.

11
Returns to scale

Decreasing returns to scale (DRS) :

• f(tz) < tf(z) for all t > 1.

12
Returns to scale

Let us check returns to scale in the Cobb-Douglas production function

β
f(z1 , z2 ) = z1α z2 . Increasing all arguments by a common factor λ, we
β
obtain f(λz1 , λz2 ) = (λz1 )α (λz2 ) β = λα+ β z1α z2 .

• When α + β = 1, we have constant returns to scale.

• When α + β > 1, we have increasing returns to scale.
• When α + β < 1, we have decreasing returns to scale.

While production functions are often homogeneous by assumption,

demand functions are homogeneous by nature (at least if we ignore the
“money illusion”.)

13
Returns to scale

Reasons for increasing returns to scale:

• larger and more advanced manufacturing facilities

• better bargaining power with suppliers and distribution channels

Reasons for decreasing returns to scale:

• management inefficiencies, resource constraints, and coordination

challenges
• creative output

14
Profit maximization problem
(PMP)
Firm’s problem: profit maximization

Assumptions:

• Firms are price takers, implying that the production plans of every
individual firm do not alter market prices p.
• This holds for both output and input.
• It implies that the firm’s share in the product and input market is
negligible.
• The production set satisfies free disposal (the producer can dispose
of the additional inputs he does not need at no cost).

We consider a very simple case where a firm using several inputs to

produce a single output.

15
Firm’s problem: profit maximization

Profit=Revenue-Cost

max pf(z) − wz
z≥0

where

• p is the price of output

• w is the vector of factor prices (rent, salary, etc.)
• z = (z1 , ..., zn ) are inputs

Note: unlike the utility maximization problem of consumers, the PMP

does not impose a budget constraint (its objective function already
considers the cost of each production plan), and thus becomes an easier
unconstrained problem.

16
Firm’s problem: profit maximization

F.O.C. of interior solutions:

∂f(z∗ )
p = wk
∂zk
∂f(z∗ )
where MPk = ∂zk .

• The market value of the marginal product obtained from using an

additional unit of input k, p · MPk must coincide with the price of
acquiring an additional unit of the input, wk .

17
Firm’s problem: profit maximization

wk
Rearranging it as p = MPk ,

wk MPk
= ≡ MRTSl,k (z∗ )
wl MPl
or

MPk MPl
=
wk wl

• The marginal productivity per dollar spent on input l is equal to that

spent on input k.

18
Firm’s problem: profit maximization

The solution of the PMP problem gives a vector of inputs z that

maximize the firm’s profits given a price vector (p,w), that is,

z(w, p) = (z1 (w, p), z2 (w, p))

• Since input demand z(w, p) is not a function of a specific production

level q, we refer to it as “unconditional factor demand”.
• Plug them into the production function to obtain the output that
arise when the firm uses its profit-maximizing input combination.

Think:
Is the first-order condition sufficient for the determination of a solution
to the PMP?

19
Firm’s problem: profit maximization

Let’s assume one input z with price w and one output f(z) = q with price
p.
Profits are given by π = pq − wz. Then on the figure below we see
q = π/p + (w/p)z for different levels of π.
• The intercept increases with π.
• The slope is fixed to w/p.
d2 f(z∗ )
When f(z) is convex: dz2
>0

20
Firm’s problem: profit maximization

d2 f ( z ∗ )
When f(z) is linear: dz2
=0

21
Firm’s problem: profit maximization

d2 f ( z ∗ )
When f(z) is concave: dz2
<0

df(z∗ ) w
The optimal point is dz = p.

22
Profit functions
Properties of profit functions

Remember the indirect utility function in consumer theory. In firm theory,

profit function is a function of (p, w) at which profit maximization
problem is solved.

• π (p, w) is homogeneous of degree 1: π (αp, αw) = απ (p, w).

• Suppose that we have access to accounting data. We observed that
all prices of input and output have scaled up by a factor t > 0.
However, profits do not scale up proportionally. We suspect that the
firm is not maximizing profits.

23
Properties of profit functions

• π (p, w) is convex in (p, w):

π (αp + (1 − α)p′ , αw + (1 − α)w′ ) ≤ απ (p, w) + (1 − α)π (p′ , w′ )

24
Properties of profit functions

• π (p, w) is convex in (p, w):

π (αp + (1 − α)p′ , αw + (1 − α)w′ ) ≤ απ (p, w) + (1 − α)π (p′ , w′ )

Proof:

π (p̄, w̄) =p̄q̄ − w̄z̄ = (αp + (1 − α)p′ )q̄ − (αw + (1 − α)w′ )z̄
=α(pq̄ − wz̄) + (1 − α)(p′ q̄ − w′ z̄)
≤απ (p, w) + (1 − α)π (p′ , w′ )

24
Properties of profit functions

π (p, w) = p · q∗ (p, w) − w · z∗ (p, w)

• Hotelling’s lemma: assume π (p, w) is differentiable at (p, w), then

∂π (p,w)
• ∂p = q∗ (p, w)
• The change in profits from a change in output price is proportional
to the quantity produced.
∂π (p,w)
• ∂wi = −zi∗ (p, w)
• The change in profits from a change in input price is proportional to
the input needed.

25
Law of supply

Combining the results in Hotelling’s lemma, and the fact that π (p, w) is
convex, its Hessian matrix
  [ ]
∂2 π ∂2 π ∂q ∂q
 ∂p
H = ∂2 π
2 ∂p∂w 
= ∂p ∂w
∂2 π ∂z ∂z
∂w∂p
− ∂p − ∂w
∂w 2

must be positive semidefinite .

Law of supply: Quantities respond in the same direction as price
changes.

• If the price of an output p increases, then the supply of the output

increases ( ∂q
∂p > 0).
• If the price of an input w increases, then the demand of the input
∂z
decreases ( ∂w < 0).

26
Law of supply

Law of demand vs. Law of supply

• In consumer theory the relationship between price and demand is

more complicated (giffen goods).
• In contrast with utility maximization, in profit maximization there is
no budget constraint.
• We have no wealth effect here, only substitution effects.

27
Cost minimization
Cost minimization

What is the difference between a cost minimizing firm and a profit

maximizing firm?

• Some “firms” do not aim at maximizing profits (say, governments,

NGOs). Instead, they target an output level (say, service) and try to
minimize their cost.
• Another way to look at the supply behavior of a firm facing
competitive output market

28
Cost minimization problem (CMP)

Consider the problem of finding a cost-minimizing way to produce

a given level of output:
min w · z
z

s.t. f(z) = q
Solve by Lagrangian:

L = w · z + λ(q − f(z))

F.O.C.:
∂f(z∗ )
wi − λ =0
∂zi
f (z∗ ) = q

29
Cost minimization

∂f(z∗ )
wi − λ =0
∂zi
∂f(z∗ )
wi ∂zi
= ∂f(z∗ )
wj
∂zj

• The left-hand side tells us at what rate factor j can be substituted

for factor i while maintaining a constant cost (the economic rate of
substitution).
• The right-hand side tells us at which rate factor j can be substituted
for factor i while maintaining a constant level of output (the
technical rate of substitution).
• They should be equal at the optimum.

30
Cost minimization

Cost minimization: the isoquant must be tangent to the constant cost

line (z2 = C/w2 − (w1 /w2 )z1 ).

31
Conditional demand function and cost function

• The optimizing set of input choices z(w, q) is known as the

conditional demand function.
• Remember that in PMP, its optimizing set of input choices z(w, p) is
“unconditional” since it is not a function of q.
• The term conditional arises because these factor demands are
conditional on the requirement that the output level q be produced.
• The optimized value of the CMP is given by the cost function
c(w, q) = w · z∗ (w, q)

32
Cost function and λ

How to understand λ?
∂f(z∗ )
1. wi − λ ∂zi =0
λ = ∂fw i
(z∗ )
∂zi

2. The Lagrange multiplier λ can be interpreted as the marginal value

∂c(w,q)
of relaxing the constraint f(z) = q. Thus, λ = ∂q , the marginal
cost of production.

33
Cost function

Properties of cost function c(w, q):

• c(.) is homogeneous of degree one in w : c(αw, q) = αc(w, q)

• c(.) is a concave function in w. (try to prove it)
∂c(w,q)
• Shepard’s lemma: ∂wi = zi (w, q).

34
Cost function

Since c(.) is a concave function in w (the Hessian matrix is negative

semidefinite ), combining with Shepard’s lemma, the element in
the diagonal of the Hessian matrix

∂2 c(w, q) ∂z (w, q)
= i ≤0
∂wi2 ∂wi

• which is consistent with the conclusions in PMP.

35
PMP and CMP: a different perspective

Remember the PMP:

max pq − w · z
zi

Using the cost function, we can restate the firm’s problem as

max pq − c(w, q)
q

F.O.C. for q∗ to be profit maximizing is then

∂c(w, q∗ ) ∂c(w, q∗ )
p− =0⇒p=
∂q ∂q
In other words, at the optimum, price equals marginal cost.

36
The geometry of cost
The geometry of cost

We denote the output by q and hold the vector of factor prices constant
at w̄ ≫ 0.

• Average cost:
C(q)
AC(q) =
q
• Marginal cost:
dC(q)
C′ (q ) =
dq

37
The geometry of cost: Case 1

A strictly decreasing returns to scale technology:

• Because of DRS, both C′ (q) and AC(q) are increasing. 38

The geometry of cost: Case 2

A constant returns to scale technology:

• Because of CRS, C′ (q) = AC(q).

39
The geometry of cost: Case 3

First increasing then decreasing returns to scale technology:

40
The geometry of cost: Case 3

Remark:
2. When MC<AC, the AC curve
decreases, and when MC>AC, the AC
curve increases.
• Intuition: using the 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 grade
1. AC=MC at q = 0, i.e., score lowers your average grade, it
must be that such new grade is lower
AC(0)=MC(0).
than your average grade thus far.

41
The geometry of cost: Case 3

Remark:
3. In other words, AC and MC curves
cross (AC=MC) at exactly the
minimum of the AC curve.
• Let us prove that when AC=MC,
∂AC(q)
∂q = 0.
∂AC(q) ∂ TC(q) q ∂TC (q)
∂q −TC(q)·1
• ∂q = ∂qq = q2
=
q·MC(q)−TC(q)
q2
= 0
• That is, q · MC(q) − TC(q) = 0 ⇒
TC(q)
MC(q) = q = AC(q).
• Hence, MC=AC at the minumum of
∂AC(q)
AC (where ∂q = 0).

42
Case 4: Variable cost and fixed cost

An important source of nonconvex cost is fixed costs.

Average cost (AC) = Average variable cost (AVC) + Average fixed cost
(AFC)

43
Short run vs. Long run

In general, the cost function can be expressed as the value of the

conditional factor demand:

c(w, q) = w · z(w, q)

The key difference between short run and long run is if there is any prior
input commitments (inputs are adjustable or not).

• In the short run, the firm does not have the flexibility of all input
choice (e.g., labor is flexible, but not capital).
• In the long run, the firm can modity the amounts of all inputs.

44
Short run vs. Long run

• In the short-run
• capital is fixed at K̄
• the firm cannot equate
MRTS with the ratio of
input prices
• In the long-run
• firm can choose input vection
A, which is a cost-minimizing
input combination.

45
Short run vs. Long run

Let zf be the vector of fixed factors, zv be the vector of variable factors.

• Short-run cost function (STC):

c(w, q, zf ) = wv zv (w, q, zf ) + wf zf

We can omit w because factor prices are taken as given,

c(q, zf ) = wv zv (q, zf ) + wf zf

• Long-run cost function (LTC):

c(w, q) = wv zv (w, q) + wf zf (w, q)

Omit w
c(q) = wv zv (q) + wf zf (q)

Note: in the short run, zf is not adjustable.

46
Short run vs. Long run: Total cost

LTC is the lower envelope of the family of short-run functions

STC generated by letting zf take all possible values.
STC = c(q, zf ) = wv zv (q, zf ) + wf zf
LTC = c(q) = wv zv (q) + wf zf (q)

47
Short run vs. Long run: Average cost

LAC is the lower envelope of the family of short-run functions SAC

generated by letting zf take all possible values.
c(q, zf ) wv zv (q, zf ) + wf zf
SAC = =
q q
c(q) wv zv (q) + wf zf (q)
LAC = =
q q

48
Aggregation in production
Is there a “representative producer”?

Consider J firms with production levels q1 , q2 , ..., qJ . Define the

aggregate supply correspondence as the sum of the individual supply
correspondences
J
q(p, w) = ∑ qj (p, w)
j=1

for j = 1, 2, ..., J.

49
Is there a “representative producer”?

There exists a representative producer:

• Producing an aggregate supply q∗ (p, w) that exactly coincides with

the sum ∑Jj=1 qj (p, w); and
• Obtaining aggregate profits π ∗ (p, w) that exactly coincide with the
sum ∑Jj=1 πj (p, w).

Intuition: The aggregate profit obtained by each firm maximizing its

profits separately (taken prices as given) is the same as that which would
be obtained if all firms were to coordinate their actions (i.e., qj′ s) in a
joint PMP.

50
Is there a “representative producer”?

• It is a “decentralization” result: to find the solution of the joint

PMP for given prices p, w, it is enough to “let each individual firm
maximize its own profits” and add the solutions of their individual
PMPs.
• Key: price taking assumptions.
• The results does not hold if firms have market power.

51
Questions?

51