1/25

Mariusz Górajski Function of several variables

 Function of several variables

Mariusz Górajski
Depratment of Econometrics, University of Lódź

April, 2016

2/25
Mariusz Górajski Function of several variables
1 Comparative statics: Envelope theorem

2 Marginal functions and Partial elasticities

3 Optimization with constraints

3/25
Mariusz Górajski Function of several variables
 Comparative statics

4/25
Mariusz Górajski Function of several variables
Comparative statics

A simple economic model can be described by

f (p, y ) = 0

where f is a function describing model, y is endogenous variable

(equilibrium), and p is a model parameter or exogenous variable.
We would like to know how the equilibrium value of y depends on
the parameter p, hence we assume that y = y (p). For example,
does it increase or decrease when the value of the parameter
increases?
Using the Implicit Function Theorem we obtain:

fp (p, y )
y 0 (p) = −
fy (p, y )

5/25
Mariusz Górajski Function of several variables
Comparative statics. Envelope theorem.
A simple optimization problem in economics can be described by

max f (x, p) (P)

x∈D

Assume that x ∗ = x ∗ (p) is the solution to (P) and

f ∗ (p) = f (x ∗ (p), p) be the value of f at x ∗ . f ∗ is the value
function for (P).
Applying the chain rule to f ∗ we obtain:
df ∗ 0
(p) = fx0 (x ∗ (p), p)x ∗ (r ) + fp0 (x ∗ (p), p).
dp

Notice that if x ∗ (p) is an interior point of D and f is differentiable

then, fx0 (x ∗ (p), p) = 0 by the Fermat theorem. Finally,

df ∗
(p) = fp0 (x ∗ (p), p). (E)
dp
6/25
Mariusz Górajski Function of several variables
Comparative statics. Envelope theorem.

Theorem (Envelope Theorem)

Let f (x, p) be a continuously differentiable function of n + m
variables. Define the value function f ∗ of m variables by

f ∗ (p) = max f (x, p),

x∈D

where x ∈ Rn and p ∈ Rm . If f ∗ is continuously differentiable then

0
fp∗j (p) = fp0j (x ∗ (p), p) (E)

for j = 1, ..., k.

7/25
Mariusz Górajski Function of several variables
Comparative statics. Envelope theorem. Example I

Example
Suppose a firm sells x units of commodity and receive revenue
R(x) = px, where p > 0 is price. The total cost function
C (x) = x 2 . Then the profit is

Π(x, p) = R(x) − C (x) = px − x 2 .

Find the optimal amounts of commodity x ∗ (p) ie. solve:

max Π(x, p) = Π∗ (p) = Π(x ∗ (p), p)

x≥0

and then determine how much profit is change if the price

increases by 1 unit from p = 4. Notice that

p p2 0 p
x ∗ (p) = Π∗ (p) = Π∗ (p) =
2 4 2
8/25
Mariusz Górajski Function of several variables
Comparative statics. Envelope theorem. Example I

Example
Using (E) we obtain the same result more simply:
0 p 0
Π∗ (p) = Π0p (x ∗ (p), p) = x ∗ (p) = Π∗ (4) = 2
2

9/25
Mariusz Górajski Function of several variables
Marginal functions and partial elasticities

10/25
Mariusz Górajski Function of several variables
Marginal products

Example
Consider the production function Y = F (K , L, T ), where Y is the
quantity of output, and K , L, S are the quantities of factor inputs
(capital, labour, land). Then

∂F ∂F ∂F
, ,
∂K ∂L ∂T
are called marginal products of capital, of labour and of land,
respectively.

11/25
Mariusz Górajski Function of several variables
Partial elasticities

Definition
A partial elasticity of f : D → R at point (x0 , y0 ) ∈ D with
respect to x is defined by
x0
Ex f (x0 , y0 ) = fx0 (x0 , y0 ) .
f (x0 , y0 )

Ex f (x0 , y0 ) is equal to the percentage change in z = f (x, y ) from

z0 = f (x0 , y0 ) caused by 1% increase in independent variable x
from x0 , when y is held constant.

12/25
Mariusz Górajski Function of several variables
Partial elasticities

Definition
A partial elasticity of f : D → R at point (x0 , y0 ) ∈ D with
respect to y is defined by
y0
Ey f (x0 , y0 ) = fy0 (x0 , y0 ) .
f (x0 , y0 )

Ey f (x0 , y0 ) is equal to the percentage change in z = f (x, y ) from

z0 = f (x0 , y0 ) caused by 1% increase in independent variable y
from y0 , when x is held constant.

The partial elasticities of the Cobb-Douglas function:

f (x, y ) = Ax a y b are constant and given by
Ex f (x, y ) = a, Ey f (x, y ) = b.

13/25
Mariusz Górajski Function of several variables
Marginal rate of substitution

In economics it is crucial to know the slope of the tangent line to a

level curve at some point.
Definition
Let z = f (x, y ) be a function of two variables. Fix a level curve
f (x, y ) = c and let y 0 be the derivative of implicit function at
(x0 , y0 ) (f (x0 , y0 ) = c). The marginal rate of substitution
MRSyx (x0 , y0 ) of y for x measures the quantity of y we must add
per unit of x removed, if the value of f is constant at c ie.

fx0 (x0 , y0 )
MRSyx (x0 , y0 ) = −y 0 = .
fy0 (x0 , y0 )

14/25
Mariusz Górajski Function of several variables
Elastistity of substitution
Definition
Let z = f (x, y ) be a function of two variables. Fix a level curve
f (x, y ) = c and let y 0 be the derivative of implicit function at
(x0 , y0 ) defined by f (x, y ) = c. The elasticity of substitution ESyx
between y and x is
y
ESyx = EMRSyx ( ).
x
If MRSyx increases by 1 %, then approximately the relative
allocation yx increases/decreases by ESyx %1 .
Example
MRS and ES for the Cobb-Douglas function f (x, y ) = Ax a y b are
given by:
by y a
MRSyx = , = MRSyx , ESyx = 1
ax x b
1
the larger the magnitude of the elasticity of substitution means the more
likely to substitute. 15/25
Mariusz Górajski Function of several variables
Optimization with constraints

16/25
Mariusz Górajski Function of several variables
Optimization problem with equality constraint
Consider a problem of maximizing or minimizing an objective
function z = f (x, y ) when x, y are restricted to satisfy an
equality constraint g (x, y ) = c.
max f (x, y ) subject to g (x, y ) = c (*)
min f (x, y ) subject to g (x, y ) = c

17/25
Mariusz Górajski Function of several variables
Optimization problem with equality constraint- example

Example (utility maximization problem)

A consumer decides about the amount of the consumption two
goods x, y using own utility function f (x, y ) = xy . She has the
budget constraint: 2x + 1y = 100, where 2 and 1 are the prices of
goods, respectively, and 100 is the consumer income. Find the only
solution to the utility maximization problem:

max xy subject to 2x + 1y = 100 (1)

18/25
Mariusz Górajski Function of several variables
The Lagrange multiplier method

We define the Lagrangian L by

L(x, y ) = f (x, y ) − λ(g (x, y ) − c), (2)

where λ ∈ R is a Lagrange multiplier.

Theorem (Lagrange’s theorem: necessary condition for (*))
Consider problem (*). Assume that f and f have continuous
partial derivatives in a domain D and that (x0 , y0 ) ∈ int(D) is a
local extreme point for f subject to g (x, y ) = c (ie. a solution to
(*)). Suppose further that gx0 (x0 , y0 ), gy0 (x0 , y0 ) are not both 0.
Then there exists a unique number λ0 such that the Lagrangian L
has stationary point at (x0 , y0 ) i.e. we have

L0x (x0 , y0 ) = 0, L0y (x0 , y0 ) = 0, g (x0 , y0 ) = c. (3)

19/25
Mariusz Górajski Function of several variables
The Lagrange multiplier method: sufficient conditions

Theorem (sufficient condition for (*))

Consider problem (*). Assume that (x0 , y0 ) is a stationary point
for the Lagrangian L and satisfies the constraint g (x0 , y0 ) = c.
Moreover,
if the Lagrangian L is concave, then (x0 , y0 ) is the global
maximum point for f subject to g (x, y ) = c.
if the Lagrangian L is convex, then (x0 , y0 ) is the global
minimum point for f subject to g (x, y ) = c.

20/25
Mariusz Górajski Function of several variables
The Lagrange multiplier method: sufficient conditions

Theorem (local second order sufficient conditions)

Consider problem (*). Assume that (x0 , y0 ) is a stationary point
for the Lagrangian L and satisfies the constraint g (x0 , y0 ) = c.
Define a border Hessian by

gx0 (x, y ) gy0 (x, y )

 
0
HL(λ, x, y ) = det  gx0 (x, y ) L00xx (x, y ) L00xy (x, y )  .
gy0 (x, y ) L00yx (x, y ) L00yy (x, y )

Then,
if HL(λ∗ , x ∗ , y ∗ ) > 0, then (x0 , y0 ) is a local maximum point
for f subject to g (x, y ) = c.
if HL(λ∗ , x ∗ , y ∗ ) < 0,, then (x0 , y0 ) is a local minimum point
for f subject to g (x, y ) = c.

21/25
Mariusz Górajski Function of several variables
The Lagrange multiplier method: example I

Consider the problem:

max xy subject to 2x + y = 100 (4)

We set f (x, y ) = xy and g (x, y ) = 2x + 1. Define the Lagrangian:

L(x, y ) = xy − λ(2x + y − 100). Then, we have L0x (x, y ) = y − 2λ,
L0y (x, y ) = x − λ and hence the necessary conditions:

 y − 2λ = 0
x −λ=0 (5)
2x + y = 100


We obtain a stationary point (x ∗ , y ∗ ) = (25, 50) of L with λ∗ = 25.

Now we examine the sufficient condition for the solution. Notice
that: L00xx (x, y ) = L00yy (x, y ) = 0 and L00xy (x, y ) = L00xy (x, y ) = 1,

22/25
Mariusz Górajski Function of several variables
The Lagrange multiplier method: example II

gx0 (x, y ) = 2, gy0 (x, y ) = 1. Hence the border Hessian HL takes

the form:
 
0 2 1
HL(λ, x, y ) = det  2 0 1  = 4 > 0.
1 1 0

Therefore (x ∗ , y ∗ ) = (25, 50) is a local maximum point for

f (x, y ) = xy st 2x + y = 100.

23/25
Mariusz Górajski Function of several variables
Interpreting λ: shadow price of resource I

max f (x, y ) subject to g (x, y ) = c (*)

Let x ∗ = x ∗ (c) and y ∗ = y ∗ (c) be the solution to (*) and

λ = λ(c) is an associated Lagrange multiplier.
The value of f at (x ∗ , y ∗ ) also depends on c:

f ∗ (c) = f (x ∗ (c), y ∗ (c))

and it is called the value function for the problem (*).

Theorem
Assume that f ∗ is differentiable in a neighbourhood of c. Then,
df ∗
(c) = λ(c)
dc
24/25
Mariusz Górajski Function of several variables
Interpreting λ: shadow price of resource II
Remark
In economics, c often denotes the available stock of some resource,
g (x, y ) represents the amount of resources used, f (x, y ) denotes
utility or profit. Then, λ(c) is called a shadow price of resource,
since it is approximately equal to change in profit or utility if one
unit of resource c is made available.

Consider the problem:

max xy subject to 2x + y = c. (6)

(x ∗ (c), y ∗ (c)) = ( c4 , c2 ) is the solution, and λ(c) = c4 .

2
Hence the value function is f ∗ (c) = m8 . It follows that
df ∗ m
dc (c) = 4 = λ(c).
λ(100) = 25 is a shadow price of income = the increase in utility
f (x, y ) = xy caused by 1 unit increase in income form c = 100.
25/25
Mariusz Górajski Function of several variables