3 Optimal choice in production

Notation: ⎫
o
y j : good j used as an output⎪
⎬
⇒ y j = y j o − y j i is the net output
i ⎪
y j : good j used as an input⎭

Net output: y = (y1 , y2 , · · · , yn ) → this is a vector. (i.e., production possibility set)

Consumption: z = (z1 , z2 , · · · , zn )

Endowment: v = (v1 , v2 , · · · , vn )
y2

T1
y2 ∗ U slope of T =slope of U ⇒ T2 =U1
U2

T (y1 , y2 ) = 0

y1 ∗ y1

max U (z) = U (z1 , z2 , . . . , zn )

y
s.t. T (y) = T (y1 , y2 , . . . , yn ) = 0

z = y+v
L = U (y + v) + λ (0 − T (y))
= U (y1 + v1 , . . . , yn + vn ) − λ · T (y1 , . . . , yn )
∂L
∂ yi = ∂U
∂ yi − λ · ∂∂ yTi = 0 ⇒ Ui = λ · Ti ⇒ UUij = Ti
Tj , ∀i, j MRS=MRT
λ = UTii if y = y(v), then λ = λ (v). If v is known, then y∗ can be solved.

However, no price (p) yet!

When there is a market, then there is p.

max U (z)
y,z
s.t. T (y) = 0
z = y + v ⇒ p′ z = p′ y + p′ v, where p′ = [ p1 , p2 , . . . , pn ]
n n n
i.e., ∑ pi z i = ∑ piyi + ∑ pivi
i=1 i=1 i=1
% &' ( % &' ( % &' (
total value o f consumption total value o f production total value o f endowment

L = U (z) + λ (0 − T (y)) + µ ( p′ y + p′ v − p′ z)
∂L ∂U pi
∂ zi = ∂ zi − µ · pi = 0, ∀i ⇒ Ui = µ pi ⇒ MRSi, j = pj

11
∂L
∂ yi = −λ ∂∂ yTi + µ · pi = 0, ∀i ⇒ λ · Ti = µ pi ⇒ Ui = λ · Ti ⇒ MRS=MRT
if m = p′ z, income=value of consumption, then maxU (z) s.t. p′ z = m

Production function

( y0 , y1 , y2 , . . . , yk ) ∈ Rn . Let y = y0 , y1 = −x1 , . . . , yn = −xn . n = k + 1

↑ % &' (
out put inputs
then (y, −x1 , . . . , −xk ) ∈ Rn , or (y, −x) ∈ Rn , x ∈ Rk
A functional relationship for (y, −x) ∈ Rn is y = f (x)
↑ ↑
out put input vector
∂ f (x)
Marginal product of xi : fi (x) = ∂ xi , ∀i = 1, . . . , k → only one input is being varied: xi
k
∂ ln y
Returns to scale (RTS) for production function: ∑ ∂ ln xi → all inputs are being varied proportionally.
i=1
x2 = K
increasing return to scale:
input doubles, output more than doubled;

input less than doubles, output doubled;

isoquant gets closer to each other as y ↑.

400
200
100
50

x1 = L

1. Cost minimization
min ∑ wi xi
x i=1
s.t. f (x1 , . . . , xn ) = y
vector notation
→
min ′
wx
x
s.t. f (x) = y
L = w′ x + λ (y − f (x))
⎧
∂L
= wi − λ fi (x) = 0, ∀i
⎪
⎪
⎨
∂ xi
⎪ ∂L
⎪
⎩ = y − f (x) = 0
∂λ
(n+1) F.O.C.’s. Solve for (x, λ )
⇒ xi = xi (w, y), ∀i → conditional factor demand (conditional on y)
λ = λ (w, y).
n
⇒ c(w, y) = ∑ wi xi (w, y) ⇔
i=1

12
c(w, y) = min w′ x
x
s.t. f (x) = y
EX: min w1 x1 + w2 x2
x
s.t. x1 x2 1−α
α =y
⇒ x1 , x2
w1 α w2 (1−α ) y
⇒ c(w, y) = → similar to e( p,U ).
(1−α )1−α α α

Properties of cost function

∂ c(w,y)
• c(w, y) is non-decreasing in w. That is, ∂w ≥0

• c(w, y) is H.O.D.=1 in w. → c(tw, y) = t · c(w, y), t > 1

k k
c(tw, y) = ∑ twi xi (w, y) = t · ∑ wi xi (w, y) = t · c(w, y)
i=1 i=1
∂ 2 c(w,y)
• c(w, y) is concave in w. That is, ∂ w2
≤ 0 for one-input case
or : c(tw + (1 − t )w′ , y) ⎡ t ) · c(w′ , y), ⎡
≥ t · c(w, y) + (1 −⎤ t ∈ [0,⎤1]
′ ′
⎢ tw1 + (1 − t )w1 ⎥ ⎢ w1 ⎥
where tw + (1 − t )w′
=⎣ ′
⎦ and w = ⎣ ⎦
′
tw2 + (1 − t )w2 w2 ′
⎡ 2
∂ 2c
⎤
∂ c
⎢ ∂ w1 2 ∂ w1 w2 ⎥
or check if ⎢
⎣ ∂ 2c
⎥ is negative definite.
∂ 2c ⎦
∂ w2 w1 ∂ w2 2
• Shepherd’s Lemma
∂ c(w,y)
xi (w, y) = ∂ wi ∀i
∂ c(w,y)
Proof: c(w, y) = w′ x ⇒ ∂ wi = xi (w, y)
∂ c(w,y)
or c(w, y) = ∑ wi xi (w, y) ⇒ ∂ wi = xi (w, y)
i
EX:
w1 α w2 1−α y ∂ c(w,y) α w1 α −1 w2 1−α y
c(w, y) = α α (1−α )1−α
⇒ x1 (w, y) = ∂ w1 = α α (1−α )1−α
= ( 1−αα )1−α ( ww21 )1−α y
∂ c(w,y)
• MC = ∂y >0

L = w′ x + λ (y − f (x)) ⇒ wi = λ fi (x), ∀i
∂L dc wi
∂y =λ = dy = MC = f i (x)
>0 → Envelope theorem.

Properties of conditional factor demand

• xi (w, y) is non-increasing in wi , ∀i
∂ c(w,y)
∂ 2 c(w,y) ∂ ∂ wi ∂ xi (w,y)
one input: ∂ wi 2
= ∂ wi = ∂ wi ≤0

13
 ∂ x1 (w,y) ∂ x1 (w,y)
EX: check ∂ w1 ≤ 0, ∂ w2 → no need.

• xi (w, y) is H.O.D=0 in w. (b/c c(w, y) is H.O.D=1 in w)

∂ 2 c(w,y) ∂ 2 c(w,y) ∂ xi (w,y) ∂ x j (w,y)
• ∂ wi ∂ w j = ∂ w j ∂ wi ⇒ ∂ w j = ∂ wi
∂ x1 (w,y) ∂ x2 (w,y)
EX: check ∂ w2 = ∂ w1

EX: check c(w, y) is concave.

use x1 = ( 1−αα )1−α ( ww21 )1−α y = k1 ( ww21 )1−α y.
x2 = ( 1−αα )α ( ww12 )α y = k2 ( w 1 α
w2 ) y.
⎡ ⎤
∂ x1 ∂ x1
⎢ ∂ w1 ∂ w2 ⎥ ∂ x1 ∂ x1 ∂ x2
∇w c = ⎢⎣ ∂ x2 ∂ x2 ⎦
⎥
∂ w1 ≤ 0. |F| = ∂ w1 ∂ w2 − ( ∂∂wx1 )2 ≥ 0.
2

∂ w1 ∂ w2
w1 α w2 1−α y
EX: obtain production function from c(w, y) = α α (1−α )1−α

use the conditional factor demand x1 and x2 .

( ww12 )α = x2 · k2 −1 y−1 ⇒ w1
w2 = x2 1/α k2 −1/α y−1/α
w1 1−α
y = x1 · k1 −1 ( w 2
) = x1 · k1 −1 (x2 1/α k2 −1/α y−1/α )1−α
1−α 1−α 1−α
= x1 · k1 −1 x2 α k2 − α y− α

1−α 1−α 1−α

⇒ y·y α = y1/α = x1 · k1 −1 x2 α k2 − α

⇒ y = x1 α x2 1−α · k1 −α k2 −(1−α ) b/c [( 1−αα )1−α ]−α · [( 1−αα )α ]−(1−α ) = 1

% &' (
1

Cost function under

⎧ CRS
⎨ y → ty
⎪
CRS: x → tx ⇒ ⇒ c ∝ y ⇒ c = c ( w) y
⎪
⎩ c → tc
see example above

2. Profit maximization

• The factor market approach

π ( p, w) = max p f (x) − w′ x
x
∂π
∂ xi = p fi (x) − wi = 0 ⇒ xi = xi ( p, w), ∀i → unconditional factor demand
n
⇒ π ( p, w) = p f (x1 ( p, w), . . . , xn ( p, w)) − ∑ wi xi ( p, w)
i=1
EX: f (x) = xα
1
∂π
∂x = p · f ′ (x) − w = 0 ⇒ p · α xα −1 = w ⇒ x( p, w) = ( pwα ) 1−α
1
⇒ π ( p, w) = pxα − wx = x( pxα −1 − w) = x( αw − w) = xw · 1−αα = w( 1−αα ) · ( pwα ) 1−α

14
 • The product market approach
π ( p, w) = max py − c(w, y)
y
∂π ∂ c(w,y)
∂y = p− ∂y =0 ⇒ y = y( p, w) → output supply function
⇒ π ( p, w) = p · y( p, w) − c(w, y( p, w))
1
EX: c(w, y) = w · y α
1
p = w · α1 · y α −1 ⇒ y = ( αwp ) 1−α
α

1 1 1 1 1
⇒ π ( p, w) = py − w · y α = y α ( p · y1− α − w) = y α ( αw − w) = w( 1−αα ) · ( αwp ) 1−α

Properties of profit function

• π ( p, w) is non-decreasing in p, non-increasing in w. b/c π ( p, w) = py − wx

• π ( p, w) is H.O.D=1 in ( p, w)

∂ 2π 2
• π ( p, w) is convex in ( p, w) i.e. ∂ p2
≥ 0, ∂∂ wπ2 ≥ 0
Graphical illustration of a profit function.

• Hotelling’s Lemma
If y( p, w) is the product supply, and xi ( p, w) is the unconditional factor demand for input i, then
∂ π ( p,w) ∂ π ( p,w)
y( p, w) = ∂p and xi ( p, w) = − ∂ wi
1
EX: π = w( 1−αα ) · ( αwp ) 1−α
1 1
y( p, w) = w( 1−αα ) · ( αw ) 1−α · ( 1−1α ) · p 1−α −1 = ( αw )−1 ( αw ) 1−α p 1−α = ( αwp ) 1−α
α α α

1 1 1
x( p, w) = −( 1−αα )(α p) 1−α ( 1−
−α
α )w
− 1−α
= ( αwp ) 1−α
⇒ y = xα the production function.

Properties of y( p, w)
∂ π ( p,w)
∂ ∂p ∂ y( p,w)
• y( p, w) is non-decreasing in p b/c ∂p = ∂p ≥ 0.

• y( p, w) is H.O.D=0 in ( p, w)

Properties of xi ( p, w)

• xi ( p, w) is non-increasing in w b/c
∂ π ( p,w)
∂ 2 π ( p,w) ∂ ∂ wi ∂ (−xi ( p,w)) ∂ xi ( p,w)
∂ wi 2
= ∂ wi = ∂ wi ≥0⇒ ∂ wi ≤ 0.

• xi ( p, w) is H.O.D=0 in ( p, w)
∂ π ( p,w) ∂ π ( p,w)
∂ ∂
∂ 2 π ( p,w) ∂ 2 π ( p,w) ∂ wi ∂wj ∂ xi ( p,w) ∂ x j ( p,w)
• ∂ wi ∂ w j = ∂ w j ∂ wi ⇒ ∂wj = ∂ wi ⇒ ∂wj = ∂ wi

15
Relationships
⎧
⎨ c(w, y) → see cost fn.
⎪
• f (x) known, then we can get
⎪
⎩ π ( p, w) → see cost fn.
⎧
⎪
⎨ f (x) → see cost fn.
• c(w, y) known, then we can get
⎪
⎩ π ( p, w) → see profit fn.

Question 5 (Varian, Exercise 5.2)

5.2. A firm has two plants with cost functions c1 (y1 ) = 3y1 2 and c2 (y2 ) = y2 2 , what is the cost
function for the firm?
Answer: If the firm has to produce a total of y = y1 + y2 (from which it follows that y2 = y − y1 ), then

total cost to produce y can then be written as

c(y) = c1 (y1 ) + c2 (y − y1 ) = 3y1 2 + (y − y1 )2 = 4y1 2 − 2yy1 + y2
since the choice of y1 must minimize c(y), we have
∂ c(y) y
∂ y1 = 8y1 − 2y = 0 ⇒ y1 = 4

Hence, upon substitution: c(y) = 34 y2

(Alternative method, producing the same result, is to equate the two MC’s.)

6.2. The cost function is c(w1 , w2 , y) = y[w1 + w2 ]. What are the conditional factor demands? What
is the production function?
Answer: By Shepherd’s Lemma
∂ c(w,y)
xi (w, y) = ∂ wi =y i = 1, 2

Hence if the firm wants to produce y units of output, it must use y = x1 = x2 units of each input regard-
less of price. There is no room for substitution.
In other words, we have a Leontief technology: y = min{x1 , x2 }

5.12. A factor of production i is called an inferior if the conditional demand for that factor de-
creases as output increases; that is ∂ xi (w, y)/∂ y < 0.

(a) Draw a diagram indicating that inferior factors are possible.

(b) Show that if the technology is constant returns to scale, then no factors can be inferior.
(c) Show that if marginal cost decreases as the price of some factor increases, then that factor must
be inferior.

16
Answer: (a) Diagram showing an inferior factor of production: x1
x2

y′
x1
x′1x1 y

(b) For constant returns to scale technology we have (Varian P.67)

c(w, y) = yc(w, 1) and thus,
∂ c(w,y) ∂ c(w,1)
xi = ∂ wi = y ∂ wi
∂ x (w,y) ∂ c(w,1)
Hence, i∂ y = ∂ wi ≥ 0 by monotonicity
∂ 2 c(w,y)
(c) we have ∂ y∂ wi <0
∂ 2 c(w,y) ∂ 2 c(w,y) ∂ xi (w,y)
Hence, by Young’s Theorem, ∂ y∂ wi = ∂ wi ∂ y = ∂y < 0.

17