Econ 2011: Calculus For Economists

Chapter-4: Functions of Several Variables

Ephrem Andargie

Addis Ababa University

[email protected]

October 4, 2023

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 1 / 92

Functions of Two Variables

Denition 4.1 (Functions of Two Variables)

A function f of two real variables x and y with domain D is a rule
that assigns a specied number f (x, y) to each point (x, y) in D.

We often write z = f (x, y) to make explicit the value taken on by f

at the general point (x, y).

The variables x and y are independent variables and z is the

dependent variable.
The domain of the function f is then the set of all possible pairs of
the independent variables, whereas the range is the set of
corresponding values of the dependent variable.

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 2 / 92

Functions of Two Variables

One way of visualizing such a function is by means of an arrow

diagram (see Figure below), where the domain is represented as a
subset of the xy -plane.

Figure

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 3 / 92

Functions of Two Variables

Example
Consider the function f that, to every pair of numbers (x, y), assigns the
number 2x + x2 y 3 . The function f is thus dened by

f (x, y) = 2x + x2 y 3

What are f (1, 0), f (0, 1), f (−2, 3), and f (a + 1, b)?

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 4 / 92

Functions of Two Variables

Example
A function of two variables appearing in many economic models is

F (x, y) = Axa y b (A, a, and b are constants)

Usually, we assume that F is dened only for x>0 and y > 0. Then F is
generally called a Cobb-Douglas function.
The Cobb-Douglas function is most often used to describe certain
production processes. Then x and y are calledinput factors, while F (x, y)
is the number of units produced, or the output. Also, F is called a
production function.

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 5 / 92

Functions of Two Variables

If a function f is given by a formula and no domain is specied, then

the domain of f is understood to be the set of all pairs (x, y) for
which the given expression is a well-dened real number.

Example
Determine the domains of the functions given by the following formulas,
xy -plane.
then draw the sets in the
√ √
(a) f (x, y) = x − 1 + y
2 p
(b) g(x, y) = 2 2 1/2
+ 9 − (x2 + y 2 )
(x + y − 4)

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 6 / 92

Geometric Representation

Points in a (three-dimensional) space can be represented by triples of

real numbers using three mutually orthogonal coordinate lines. Figure

The graph of a function f of two variables is a surface in a

three-dimensional space with equation z = f (x, y). Figure

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 7 / 92

Geometric Representation

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 8 / 92

Geometric Representation

¦
¥ £
Í

ð
Ü §
¨

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 9 / 92

Level Curves

A second kind of geometric representation, borrowed from mapmakers,

are level curves.

Denition 4.2
The level curves of a function f of two variables are the curves with
equations f (x, y) = k , where k is a constant (in the range of f ).

A level curve f (x, y) = k is the set of all points in the domain of f at

which f takes on a given value k. Figure

In other words, it shows where the graph of f has height k.

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 10 / 92

Level Curves

¦
ìë

¨ §
µãìë
µãìð
µãíë
µãíð
µãîë
µãîð

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 11 / 92

Level Curves

Example
Consider the function of two variables dened by the equation

z = x2 + y 2 (1)

What are the level curves? Draw a set of level curves in the xy -plane.

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 12 / 92

Level Curves

Example
Consider the function of two variables dened by the equation

z = x2 + y 2 (1)

What are the level curves? Draw a set of level curves in the xy -plane.

Solution
The variable z can only assume values ≥ 0. Each level curve has the
equation
x2 + y 2 = k
for some k ≥ 0. We see that these are circles in the
√ xy -plane centered at
the origin and with radius k. Figure

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 12 / 92

Level Curves

§ ¦
½k ë
½k ì
½k í
½k î
½k ï

§
¨

(a) Level curves of z = x2 + y2 (b) The graph of z = x2 + y2

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 13 / 92

Level Curves

Example
Suppose F (K, L) denotes a rm's output, where K denotes capital input
and L denotes labor input. A level curve for the function is a curve in the
KL-plane given by

F (K, L) = Y0 (Y0 is a constant)

This curve is called an isoquant (signifying equal quantity. For a

Cobb-Douglas function F (K, L) = AK a Lb with a+b<1 and A > 0,
Figures show a part of the graph near the origin, and three of the
isoquants.

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 14 / 92

Level Curves

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 15 / 92

Partial Derivatives

Given f (x, y), suppose we let only x vary while keeping y xed, say
y = b, where b is a constant.

Then we are really considering a function of a single variable x,

namely g(x) = f (x, b).
If g has a derivative at a, then we call it the partial derivative of f
w.r.t. x at (a, b) and denoted it by fx (a, b).
By the denition of the derivative, we have

g(a + h) − g(a)
fx (a, b) = g ′ (a) = lim
h→0 h
f (a + h, b) − f (a, b)
= lim (2)
h→0 h

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 16 / 92

Partial Derivatives

Similarly, the partial derivative of f w.r.t.y at (a, b), denoted by

fy (a, b), is obtained by keeping x xed (x = a) and nding the
ordinary derivative at b of the function h(y) = f (a, y):

f (a, b + h) − f (a, b)
fy (a, b) = lim (3)
h→0 h
If we now let the point vary in (2) and (3), fx and fy become
functions of two variables.

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 17 / 92

Partial Derivatives

Iff is a function of two variables, its partial derivatives are the functions
fx and fy dened by

f (x + h, y) − f (x, y)
fx (x, y) = lim
h→0 h
f (x, y + h) − f (x, y)
fy (x, y) = lim
h→0 h

Alternative notations for the partial derivative: If z = f (x, y), we write

∂f ∂ ∂z
fx (x, y) = fx = = f (x, y) = = f1
∂x ∂x ∂x
∂f ∂ ∂z
fy (x, y) = fy = = f (x, y) = = f2
∂y ∂y ∂y

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 18 / 92

Partial Derivatives

Rule for Finding Partial Derivatives of z = f (x, y)

1 To nd fx , regard y as a constant and dierentiate f (x, y) w.r.t. x.
2 To nd fy , regard x as a constant and dierentiate f (x, y) w.r.t. y.

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 19 / 92

Partial Derivatives

Rule for Finding Partial Derivatives of z = f (x, y)

1 To nd fx , regard y as a constant and dierentiate f (x, y) w.r.t. x.
2 To nd fy , regard x as a constant and dierentiate f (x, y) w.r.t. y.

Example
1 If f (x, y) = x3 + x2 y 3 − 2y 2 , nd fx (2, 1) and fy (2, 1).
2 Find the partial derivatives of the following functions:

(a) f (x, y) = x3 y + x2 y 2 + x + y 2
xy
(b) f (x, y) = x2 +y 2

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 19 / 92

Geometric Interpretations of Partial Derivatives

Let z = f (x, y) be a function of two variables, with its graph as shown

in Figure .

The partial derivative fx (x0 , y0 ) is the derivative of z = f (x, y0 ) w.r.t.

x at the point x = x0 , and is therefore the slope of the tangent line ly
to the curve Ky at x = x0 .
In the same way, fy (x0 , y0 ) is the slope of the tangent line lx to the
curve Kx at y = y0 .

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 20 / 92

Geometric Interpretations of Partial Derivatives

Figure: The partial derivatives of f at (x0 , y0 ) are the slopes of the tangents to
Kx and Ky .

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 21 / 92

Higher-Order Partial Derivatives

If z = f (x, y), then ∂f /∂x and ∂f /∂y are called rst-order partial
derivatives.
These partial derivatives are, in general, again functions of two
variables.
From ∂f /∂x, provided this derivative is itself dierentiable, we can get
two new functions by taking the partial derivatives w.r.t. x and y .
In the same way, we can take the partial derivatives of ∂f /∂y w.r.t. x
and y .
The four functions we obtain by dierentiating twice in this way are
called second-order partial derivatives of f (x, y).

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 22 / 92

Higher-Order Partial Derivatives

If z = f (x, y), we use the following notation:

∂2f ∂2z
 
∂ ∂f
(fx )x = fxx = f11 = = =
∂x ∂x ∂x2 ∂x2
∂2f ∂2z
 
∂ ∂f
(fx )y = fxy = f12 = = =
∂y ∂x ∂y∂x ∂y∂x
∂2f ∂2z
 
∂ ∂f
(fy )x = fyx = f21 = = =
∂x ∂y ∂x∂y ∂x∂y
∂2f ∂2z
 
∂ ∂f
(fy )y = fyy = f22 = = =
∂y ∂y ∂y 2 ∂y 2

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 23 / 92

Higher-Order Partial Derivatives

Example
Find the second partial derivative of

f (x, y) = x3 + x2 y 3 − 2y 2

Notice that fxy = fyx in the above example. This is not just a
coincidence.

It turns out that the mixed second-order partial derivatives (or

cross-partials) are usually equal. That is, for most functions
z = f (x, y), it is the case that

∂2f ∂2f
=
∂x∂y ∂y∂x

Sucient conditions for this equality are given in Theorem 4.1

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 24 / 92

Functions of More Variables

Any ordered collection of n numbers (x1 , x2 , . . . , xn ) is called an

n-vector.
n-vectors are often denoted by bold letters. For example, we write
x = (x1 , x2 , . . . , xn ).

Denition 4.3 (Functions of n Variables)

Given any set D of n-vectors, a function f of n variables
(x1 , x2 , . . . , xn ) with domain D is a rule that assigns a specied
number f (x) = f (x1 , x2 , . . . , xn ) to each n-vector x = (x1 , x2 , . . . , xn ) in
D.

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 25 / 92

Euclidean n-Dimensional Space

The set of all possible n-vectors x = (x1 , x2 , . . . , xn ) of real numbers

is usually called the Euclidean n-dimensional space, or n-space,
and is denoted by Rn .
More formally,

Rn = |R × R ×{z· · · × R}
n−times
= {(x1 , x2 , . . . , xn ) : x1 ∈ R, x2 ∈ R, . . . , xn ∈ R},

where × denotes the Cartesian product.

For n = 1, 2, and 3, we have geometric interpretations of Rn as a

line, a plane, and a three-dimensional space, respectively.

But for n ≥ 4, there is no geometric interpretation.

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 26 / 92

Euclidean n-Dimensional Space

z = f (x1 , x2 , . . . , xn ) = f (x) represents a function of n variables,

If
we dene the graph of f as the set of all points (x, f (x)) in Rn+1 for
which x belongs to the domain of f .
We also call the graph a surface (or sometimes a hypersurface) in
Rn+1 .
For z = z0 (a constant), the set of points in Rn satisfying f (x) = z0
is called a level surface of f .

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 27 / 92

Euclidean n-Dimensional Space

z = f (x1 , x2 , . . . , xn ) = f (x) represents a function of n variables,

If
we dene the graph of f as the set of all points (x, f (x)) in Rn+1 for
which x belongs to the domain of f .
We also call the graph a surface (or sometimes a hypersurface) in
Rn+1 .
For z = z0 (a constant), the set of points in Rn satisfying f (x) = z0
is called a level surface of f .
Example
Let f (x, y, z) = xy + xz + yz . Find f (−2, 1, 3) and
f (a + 1, b + 1, c + 1) − f (a, b, c).
Show that f (tx, ty, tz) = t2 f (x, y, z).

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 27 / 92

Partial Derivatives with More Variables

Ifz = f (x) = f (x1 , x2 , . . . , xn ), its partial derivative w.r.t. the ith

variable xi is
∂z f (x1 , . . . , xi + h, . . . , xn ) − f (x1 , . . . , xi , . . . , xn )
= lim
∂xi h→0 h
and we also write
∂z ∂f
= = fxi = fi
∂xi ∂xi

Example
Find f1 , f2 , and f3 if f (x1 , x2 , x3 ) = 5x21 + x1 x32 − x22 x23 + x33 .

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 28 / 92

Partial Derivatives with More Variables
For each of the n rst-order partials o,we have n second-order
partials:
∂2f
 
∂ ∂f
= = fij
∂xj ∂xi ∂xj ∂xi
Here both i and j may take any value 1, 2, . . . , n, so altogether there
2
are n second-order partials.

We usually display these second-order partials in an n×n square array

as follows

f11 (x) f12 (x) · · ·


f1n (x)


 f21 (x) f22 (x) · · · f2n (x) 

 
H(x) =  .

. .. .

 (The Hessian) (4)
 .. .
. . .
.

 
fn1 (x) fn2 (x) · · · fnn (x)

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 29 / 92

Partial Derivatives with More Variables

(4) is called the Hessian matrix of f at x = (x1 , x2 , . . . , xn ).

The n second-order partial derivatives fii found by dierentiating
twice w.r.t. the same variable are called direct second-order partials.
The others, fij where i ̸= j , are mixed or cross partials.

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 30 / 92

Partial Derivatives with More Variables

(4) is called the Hessian matrix of f at x = (x1 , x2 , . . . , xn ).

The n second-order partial derivatives fii found by dierentiating
twice w.r.t. the same variable are called direct second-order partials.
The others, fij where i ̸= j , are mixed or cross partials.

Example
Find the Hessian matrix of f (x1 , x2 , x3 ) = 5x21 + x1 x32 − x22 x23 + x33 .

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 30 / 92

Young's Theorem

If z = f (x1 , x2 , . . . , xn ), then the two second-order cross-partial

derivatives fij and fji are usually equal. That is,
   
∂ ∂f ∂ ∂f
=
∂xj ∂xi ∂xi ∂xj

As the following theorem indicates, the order of dierentiation is

irrelevant in the determination of cross-partial derivatives.

Theorem 4.1 (Young's Theorem)

For a function z = f (x1 , x2 , . . . , xn ), with continuous rst- and
second-order partial derivatives, the order of dierentiation in computing
the cross-partials is irrelevant. That is, fij = fji for any pair i, j ;
i, j = 1, 2, . . . , n; i ̸= j . (Of course, this statement holds trivially when
i = j as well.)

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 31 / 92

Dierentials

Consider a dierentiable function one variable y = f (x), and let dx

denote an arbitrary change in the variable x.
′
The expression f (x) dx is called the dierential of y = f (x), and it is
denoted by dy (or df ), so that

dy = f ′ (x) dx (5)

We dene the dierential dx to be an independent variable; that is, dx

can be given the value of any real number.

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 32 / 92

Dierentials

Figure shows the relationship between the increment ∆y and the

dierential dy :
∆y represents the change in height of the curve:
∆y = f (x + dx) − f (x).
dy represents the change in height of the tangent line when x changes
by an amount dx = ∆x.
As a result, we have ∆y ≈ dy = f ′ (x) dx.

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 33 / 92

Dierentials

Figure: A geometric representation of the dierential dy and

∆y = f (x + dx) − f (x)

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 34 / 92

Dierentials

Example
√
Find the dierential dy , given f (x) = x + 3.

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 35 / 92

Dierentials

Example
√
Find the dierential dy , given f (x) = x + 3.

Solution
dx
dy = f ′ (x)dx = √ □
2 x+3

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 35 / 92

Dierentials

Example
√
Find the dierential dy , given f (x) = x + 3.

Solution
dx
dy = f ′ (x)dx = √ □
2 x+3

Example
Compare the values of ∆y and dy if y = f (x) = x3 + x2 − 2x + 1 and x
changes (a) from 2 to 2.05 and (b) from 2 to 2.01.

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 35 / 92

Dierentials

Solution
(a) We have

f (2) = 23 + 22 − 2(2) + 1 = 9
f (2.05) = (2.05)3 + (2.05)2 − 2(2.05) + 1 = 9.717625
∆y = f (2.05) − f (2) = 0.717625

In general,
dy = f ′ (x)dx = (3x2 + 2x − 2)dx
When x=2 and dx = ∆x = 0.05, this becomes

dy = [3(2)2 + 2(2) − 2]0.05 = 0.7

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 36 / 92

Dierentials

Solution (continued )
(b)

f (2.01) = (2.01)3 + (2.01)2 − 2(2.01) + 1 = 9.140701

∆y = f (2.01) − f (2) = 0.140701

When dx = ∆x = 0.01,

dy = [3(2)2 + 2(2) − 2]0.01 = 0.14

Notice that the approximation ∆y ≈ dy becomes better as ∆x becomes

smaller.

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 37 / 92

Dierentials

For a dierentiable function of two variables, z = f (x, y), we dene

the dierentials dx and dy to be independent variables; that is, they
can be given any values.

Then the dierential dz (sometimes denoted df ), also called the

total dierential, is dened by
∂z ∂z
dz = fx (x, y)dx + fy (x, y)dy = dx + dy (6)
∂x ∂y

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 38 / 92

Dierentials

For a dierentiable function of two variables, z = f (x, y), we dene

the dierentials dx and dy to be independent variables; that is, they
can be given any values.

Then the dierential dz (sometimes denoted df ), also called the

total dierential, is dened by
∂z ∂z
dz = fx (x, y)dx + fy (x, y)dy = dx + dy (6)
∂x ∂y

Example
(a) If z = f (x, y) = x2 + 3xy − y 2 , nd the dierential dz .
(b) If x changes from 2 to2.05 and y changes from 3 to 2.96, compare
the values of ∆z and dz .

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 38 / 92

Dierentials
Total Dierentials

Solution
(a) (6) gives

∂z ∂z
dz = dx + dy = (2x + 3y)dx + (3x − 2y)dy
∂x ∂y

(b) Putting x = 2, dx = ∆x = 0.05, y = 3, and dy = ∆y = −0.04, we

get
dz = [2(2) + 3(3)]0.05 + [3(2) − 2(3)](−0.04) = 0.65
The increment of z is

∆z = f (2.05, 2.96) − f (2, 3)

= [(2.05)2 + 3(2.05)(2.96) − (2.96)2 ] − [22 + 3(2)(3) − 32 ]
= 0.6449 □

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 39 / 92

Dierentials

For the case of a function of n independent variables given by the

general form
y = f (x1 , x2 , . . . , xn )
the total dierential of this function can be written as

∂f ∂f ∂f
dy = dx1 + dx2 + · · · + dxn
∂x1 ∂x2 ∂xn
n
X ∂f
= dxi (7)
∂xi
i=1

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 40 / 92

Second-Order Dierentials

The second-order dierential is dened as the dierential of a

dierential.

For a function of one variable, y = f (x), the second-order dierential

is given as d(dy), commonly denoted by d2 y .
Given that dy = f ′ (x) dx, we can obtain d2 y by further dierentiation
of dy , where dx is to be treated as a constant during dierentiation.
(Why?)

Hence, dy can in the nal analysis vary only with x. In view of this,
we have

d2 y ≡ d(dy) = d[f ′ (x) dx]

= [df ′ (x)] dx
= [f ′′ (x) dx] dx = f ′′ (x) dx2

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 41 / 92

Second-Order Dierentials
Second-Order Total Dierential for a Function of Two Variables

Given the total dierential (6) of z = f (x, y), we can derive an

expression for the second-order total dierential d2 z by further
dierentiation of dz .
In doing so, we should remember that in the equation
dz = fx dx + fy dy , the symbols dx and dy represent arbitrary or given
changes in x and y ; so they must be treated as constants during
dierentiation.

As a result, dz depends only on fx and fy and since fx and fy are

themselves functions of x and y , dz , like z itself, is a function of x
and y.

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 42 / 92

Second-Order Dierentials
Second-Order Total Dierential for a Function of Two Variables

To obtain d2 z , we merely apply the denition of a dierential to dz

itself. Thus,

∂(dz) ∂(dz)
d2 z = d(dz) = dx + dy
∂x ∂y
∂ ∂
= (fx dx + fy dy)dx + (fx dx + fy dy)dy
∂x ∂y
= (fxx dx + fyx dy)dx + (fxy dx + fyy dy)dy
= fxx dx2 + fyx dydx + fxy dxdy + fyy dy 2
= fxx dx2 + 2fxy dxdy + fyy dy 2 [fxy = fyx ] (8)

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 43 / 92

Second-Order Dierentials (Optional)
Second-Order Total Dierential for a Function of n Variables

Recall that the rst-order total dierential of y = f (x1 , . . . , xn )

becomes

n
X
dy = f1 dx1 + f2 dx2 + · · · + fn dxn = fi dxi
i=1

Since each of the partial derivatives fi is a function of

(x1 , x2 , . . . , xn ), dy is also a function of (x1 , x2 , . . . , xn ), and so we
can nd its total dierential d[dy] or d2 y as follows:

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 44 / 92

Second-Order Dierentials (Optional)
Second-Order Total Dierential for a Function of n Variables

∂[dy] ∂[dy] ∂[dy]

d2 y = dx1 + dx2 + · · · + dxn
∂x1 ∂x2 ∂xn
Pn Pn
∂ [ i=1 fi dxi ] ∂ [ i=1 fi dxi ]
= dx1 + dx2 + · · ·
∂x ∂x2
Pn1
∂ [ i=1 fi dxi ]
+ dxn (*)
∂xn
Each term in (6) represents n partial derivatives. For example,
Pn
∂ [ i=1 fi dxi ] ∂ [f1 dx1 + f2 dx2 + · · · + fn dxn ]
=
∂x1 ∂x1
∂f1 ∂f2 ∂fn
= dx1 + dx2 + · · · + dxn
∂x1 ∂x1 ∂x1
= f11 dx1 + f21 dx2 + · · · + fn1 dxn

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 45 / 92

Second-Order Dierentials (Optional)
Second-Order Total Dierential for a Function of n Variables

It follows that since each of the n terms in (*) represents n terms,

there are n2 terms altogether. It is tedious to write these out in full.
However, by using the sigma notation, we get

n  n
n X
∂ [ ni=1 fi dxi ]
X P  X
2
d y= dxj = fij dxi dxj
∂xj
j=1 j=1 i=1

In some applications, it is useful to write d2 y in matrix notation

d2 y = dxT Hdx

dxT = dx1 dx2

 
where · · · dxn is the vector of changes dxi and
H is the Hessian matrix given in (4).

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 46 / 92

Optimization
The Dierential Version of Optimization Conditions

Given a function y = f (x), we write the dierential

dy = f ′ (x) dx

Since points where f ′ (x) ̸= 0 and hence dy ̸= 0 (for a nonzero dx)

cannot qualify as stationary points, it follows that a necessary
condition for z to attain an extremum is dy = 0.
Thus the rst-order derivative condition  f
′ (x)
= 0 can be translated
into the rst-order dierential condition  dy = 0 for an arbitrary
nonzero dx.

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 47 / 92

Optimization
The Dierential Version of Optimization Conditions

Next, we derive the dierential equivalents of second-order sucient

conditions.

Recall that the second-order dierential of y = f (x) is given as

d2 y = f ′′ (x) dx2

Since we are considering nonzero values of dx only, the dx2 term is

2
always positive; thus d y and f ′′ (x) must take the same sign.
It follows that the derivative condition  f
′′ (x) <0 is sucient for a
maximum of y can equivalently be stated as the dierential condition
d
2y <0 for an arbitrary nonzero dx is sucient for a maximum of y .
The translation of condition for a minimum of y is analogous.

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 48 / 92

Optimization
Extreme Values of a Function of Two Variables

Geometrically, for a function of two choice variables, the graph of the

functionz = f (x, y)becomes a surface in 3-space, and the extreme
values are associated with peaks and bottoms of these surfaces.

They will, in this new context, be shaped like domes and bowls,
respectively. (See Figure )

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 49 / 92

Optimization
Extreme Values of a Function of Two Variables

Figure

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 50 / 92

Optimization
Extreme Values of a Function of Two Variables: First-Order Condition

For the function z = f (x, y) the rst-order necessary condition for an

extremum (either maximum or minimum) again involves dz = 0.
But since there are two independent variables here, dz is now a total
dierential; thus the rst-order condition should be modied to the
form

dz = 0 for arbitrary value of dx and dy, not both zero (9)

The rationale behind this condition is similar to the explanation of the

condition dz = 0 for the one-variable case: an extremum point must
be a stationary point, and at a stationary point, dz as an
approximation to the actual change ∆z must be zero for arbitrary dx
and dy , not both zero.

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 51 / 92

Optimization
Extreme Values of a Function of Two Variables: First-Order Condition

In the present two-variable case, the total dierential is

dz = fx dx + fy dy

In order to satisfy condition (9), the two partial derivatives fx and fy

must simultaneously be equal to zero.

Thus the equivalent derivative version of the rst-order condition (9) is

fx = fy = 0 (10)

A point (x0 , y0 ) is called a critical point (or stationary point ) of f if

fx (x0 , y0 ) = 0 and fy (x0 , y0 ) = 0.

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 52 / 92

Optimization
Extreme Values of a Function of Two Variables: First-Order Condition

There is a simple graphical interpretation of the condition in (10).

With reference to point A in Figure: (a) , to have fx = 0 at that point

means that the tangent line Tx , drawn through A and parallel to the
xz plane (holding y constant), must have a zero slope.

By the same toke to have, fy = 0 at point A means that the tangent

line Ty , drawn through A and parallel to the yz plane (holding x)
constant, must have a zero slope.

It can be veried that these tangent-line requirements actually also

apply to the minimum point B in Figure: (b) .

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 53 / 92

Optimization
Extreme Values of a Function of Two Variables: First-Order Condition

The rst-order condition is necessary, but not sucient. That it is not

sucient to establish an extremum can be seen from Figure .

At point C, both Tx and Ty have zero slopes, but this point does not
qualify as an extremum: Where as it is a minimum when viewed
against the background of the yz plane, it turns out to be a maximum
when looked at against the xz plan.

A point with such dual personality is referred to as a saddle point.

To develop a sucient condition, we must look to the second-order
total dierential, which is related to second-order partial derivatives.

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 54 / 92

Optimization
Extreme Values of a Function of Two Variables: First-Order Condition

Figure

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 55 / 92

Optimization
Extreme Values of a Function of Two Variables: Second-Order Conditions

Suppose that (x0 , y0 ) is a stationary point of the function f.

If we make small changes in any direction from this point, and the
result is to reduce the value of f, then this point must yield a local
maximum of the function.

Similarly, if we move away from this point a small distance in any

direction and this increases the value of the function, this point must
yield a local minimum.

Finally, if moving in some directions increases the value of the

function, while moving in other directions reduces the function value,
then the stationary point must be a saddle point.

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 56 / 92

Optimization
Extreme Values of a Function of Two Variables: Second-Order Conditions

Given z = f (x, y) with continuous partial derivatives, its total

dierential dz for arbitrary changes dx and dy gives the change in the
function in an arbitrary direction.

Suppose this function is the total dierential dz . At the stationary

point (x0 , y0 ), this total dierential dz = df (x, y) = 0.
If, for any (small) movement away from this point dz becomes
negative, that means the function is decreasing and (x0 , y0 ) yields a
maximum.
If, for any (small) movement away from this point dz becomes
positive, that means the function value is increasing and (x0 , y0 ) yields
a minimum.
Thus sucient conditions for a local maximum or minimum can be
expressed in terms of what happens to dz as we move away from
(x0 , y0 ) in any direction, namely in terms of the second-order
dierential d2 z .
Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 57 / 92
Optimization
Extreme Values of a Function of Two Variables: Second-Order Conditions

Thus, once the rst-order necessary condition is satised, the

second-order sucient condition for a maximum of z = f (x, y) is

d2 z < 0 for arbitrary values of dx and dy, not both zero

A positive d2 z value at a stationary point, on the other hand, is

associated with a local minimum. The second-order sucient
condition for a minimum of z = f (x, y) is

d2 z > 0 for arbitrary values of dx and dy, not both zero

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 58 / 92

Optimization
Extreme Values of a Function of Two Variables: Second-Order Conditions

To determine the sign of the second-order total dierential d2 z in

terms of signs of the second-order partial derivatives we need to
perform completing the square on the expression for d2 z .
By adding
2 dy 2 /f
fxy to, and subtracting the same quantity from,
xx
the right side of (8), we can rewrite d2 z as follows:

2
fxy 2
fxy
d2 z = fxx dx2 + 2fxy dxdy + dy 2 + fyy dy 2 − dy 2
fxx fxx
! !
f 2 f 2
2f xy xy xy
= fxx dx2 + dxdy + 2 dy 2 + fyy − dy 2
fxx fxx fxx
2 2
fxx fyy − fxy

fxy
= fxx dx + dy + (dy 2 )
fxx fxx

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 59 / 92

Optimization
Extreme Values of a Function of Two Variables: Second-Order Conditions

Now that the variables dx and dy appear only in squares, we can

predict the sign of d2 z entirely on the values of the coecients fxx ,
fyy , and fxy as follows:

d2 z > 0 ⇐⇒ fxx > 0 and

2
fxx fyy − fxy >0
d2 z < 0 ⇐⇒ fxx < 0 and
2
fxx fyy − fxy >0

Now that (i)

2
fxx fyy − fxy should be positive in both cases and (ii) as
2
a prerequisite for the positivity of fxx fyy − fxy , the product fxx fyy
2
must be positive (since it must exceed the squared term fxy ), hence,
this condition automatically implies that fxx and fyy must take
identical algebraic signs.

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 60 / 92

Optimization
Extreme Values of a Function of Two Variables: Second-Order Conditions

In summary, the necessary (rst-order) conditions for an extremum of

a function of two variables z = f (x, y) are

fx = fy = 0

and the sucient (second-order) conditions are

2
fxx > 0 and fxx fyy − fxy >0 for a local minimum
2
fxx < 0 and fxx fyy − fxy >0 for a local maximum

If, on the other hand,

2 < 0,
fxx fyy − fxy the stationary point yields a
saddle point.

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 61 / 92

Optimization
Extreme Values of a Function of Two Variables: Second-Order Conditions

Example
Find stationary values of f (x, y) = x2 + y 2 − 2x − 6y + 14, and determine
whether each stationary point is a local maximum, minimum, or saddle
point.

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 62 / 92

Optimization
Extreme Values of a Function of Two Variables: Second-Order Conditions

Example
Find stationary values of f (x, y) = x2 + y 2 − 2x − 6y + 14, and determine
whether each stationary point is a local maximum, minimum, or saddle
point.

Solution
The rst-order conditions are

fx = 2x − 2 = 0
fy = 2y − 6 = 0

These partial derivatives are equal to 0 when x=1 and y = 3, so the only
critical point is (1, 3).

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 62 / 92

Optimization
Extreme Values of a Function of Two Variables: Second-Order Conditions

Solution (continued )
The second-order partial derivatives are

fxx = 2 fyy = 2 and fxy = fyx = 0

Since fxx > 0 and fxx fyy − fxy2 = 4 > 0 for all x and y , the stationary

point (1, 3) yields a unique minimum of the function and the minimum of
the function is

f (1, 3) = 12 + 32 − 2(1) − 6(3) + 14 = 4.

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 63 / 92

Using Dierentials to Find the Gradient of Level Curves

Recall that we can represent a function z = f (x, y) by drawing the

level curves: the lines f (x, y) = k for dierent values of the constant
k, as shown in Figure .

To nd the gradient at the point A, consider a small change in x and

y that takes you to a point B on the same level curve.

Then the change in z is given by:

∂z ∂z
dz = dx + dy
∂x ∂y

But in moving from A to B, z does not change, so dz = 0:

∂z ∂z
dx + dy = 0
∂x ∂y

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 64 / 92

Using Dierentials to Find the Gradient of Level Curves

¦
ìë

¨ §
µãìë
µãìð
µãíë
µãíð
µãîë
µãîð

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 65 / 92

Using Dierentials to Find the Gradient of Level Curves

dy
Rearranging this equation gives us
dx :

The gradient of the level curve of the function z = f (x, y) at any point
(x, y) is given by:
∂z
dy fx
= − ∂x
∂z
=−
dx ∂y
fy

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 66 / 92

Using Dierentials to Find the Gradient of Level Curves

dy
Rearranging this equation gives us
dx :

The gradient of the level curve of the function z = f (x, y) at any point
(x, y) is given by:
∂z
dy fx
= − ∂x
∂z
=−
dx ∂y
fy

Example
The above Figure shows the function z = x2 + 2y 2 .
(i) Which level curve passes through the point where x=2 and y = 1?
At (2, 1), z = 6, so the level curve is x2 + 2y 2 = 6.

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 66 / 92

Using Dierentials to Find the Gradient of Level Curves

Example
(ii) Find the gradient of the isoquant at this point.

∂z ∂z
dz = dx + dy = 0
∂x ∂y
=⇒ 2x dx + 4y dy = 0
dy x
=⇒ =−
dx 2y

So at (2, 1) the gradient is −1.

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 67 / 92

Constrained Optimization

In many economic problems there exist one or more constraints which

restrict the set of x-values we are allowed to consider as possible
solutions.

In this section we develop techniques for dealing with a type of

constraint in which the x-variables are restricted to a set of values
that satisfy one or more functional equations.

The main topic is the derivation and application of the method of

Lagrange multipliers.

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 68 / 92

Constrained Optimization
Constrained Problems and Approaches to Solutions

Suppose that we wish to maximize a function f (x1 , x2 ), where f is

strictly concave, continuous and twice dierentiable.

In the absence of constraints, we know from our forgoing discussion

that the solution (x∗1 , x∗2 ) satises the conditions fi (x∗1 , x∗2 ) = 0,
i = 1, 2.
Now suppose that we impose the constraint

g(x1 , x2 ) = 0

where g is continuous and twice dierentiable also.

This means that we are only allowed to consider as possible solutions

to the problem those pairs of (x1 , x2 )-values which satisfy that
equation.

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 69 / 92

Constrained Optimization
Constrained Problems and Approaches to Solutions

Figure

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 70 / 92

Constrained Optimization
Constrained Problems and Approaches to Solutions

In Figure the function f, is illustrated (a) in three dimensions and (b)

by the associated level curves.

The curve G in each part of the gure represents the set of

(x1 , x2 )-pairs that satisfy a constraint such as the one above.
The problem of maximizing the function subject to the constraint is,
stated diagrammatically, the problem of nding a point on the curve
G which gives the highest possible value of the function.

The answer can be seen in Figure: (a) , but it emerges more clearly in
Figure: (b) .

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 71 / 92

Constrained Optimization
Constrained Problems and Approaches to Solutions

The level curve labeled f∗ is clearly the highest level curve of f that
can be reached while remaining on the curve G, and so the point
(x∗1 , x∗2 ) is the solution to the problem: it gives the highest value of
the function consistent with satisfying the constraint.

The most obvious feature about the point (x∗1 , x∗2 ) is that it is a point
of tangency between the curve G and the level curve f ∗.
Given the shapes of the level curves of f and the curve G, it should be
clear that any solution to the problem must be at a point of tangency.
This could be tested by shifting the curve G slightly while keeping its
general shape.

Suppose now that we want to express this diagrammatic solution

algebraicallythat would certainly simplify the problem of nding the
numerical values of the solution.

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 72 / 92

Constrained Optimization
Constrained Problems and Approaches to Solutions

We know that we can write the slope of a level curve of f (x1 , x2 ) as

dx2 f1 (x1 , x2 )
=−
dx1 f2 (x1 , x2 )

and the slope of the curve G as

dx2 g1 (x1 , x2 )
=−
dx1 g2 (x1 , x2 )

Since at a point of tangency these slopes must be equal, we have that

at (x∗1 , x∗2 ),
f1 (x∗1 , x∗2 ) g1 (x∗1 , x∗2 )
= (11)
f2 (x∗1 , x∗2 ) g2 (x∗1 , x∗2 )

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 73 / 92

Constrained Optimization
Constrained Problems and Approaches to Solutions

Note that this is only one equation in two unknowns and so in itself
cannot give us a solution for the values x∗1 and x∗2 .
The second equation we need is given by the constraint

g(x∗1 , x∗2 ) = 0 (12)

since if x∗1 and x∗2 are optimal, then they must be feasible and so
satisfy the constraint.

(11) and (12) taken together allow us to solve for the values x∗1 and
x∗2 .

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 74 / 92

Constrained Optimization
Constrained Problems and Approaches to Solutions

We now set out the Lagrange multiplier technique for solving

constrained optimization problems, and we shall justify it by showing
that it is equivalent to the two approaches we have just examined.

We proceed by introducing a new variable,λ, the Lagrange

multiplier, and by forming the Lagrange function or Lagrangean
L(x1 , x2 , λ) = f (x1 , x2 ) + λg(x1 , x2 )

We then nd a stationary point of L with respect to x1 , x2 , and λ.

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 75 / 92

Constrained Optimization
Constrained Problems and Approaches to Solutions

This gives the conditions

∂L
= f1 (x∗1 , x∗2 ) + λ∗ g1 (x∗1 , x∗2 ) = 0
∂x1
∂L
= f2 (x∗1 , x∗2 ) + λ∗ g2 (x∗1 , x∗2 ) = 0
∂x2
∂L
= g(x∗1 , x∗2 ) = 0
∂λ
We have the three equations to solve for the three unknowns, x∗1 , x∗2 ,
and λ∗ .

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 76 / 92

Constrained Optimization
Constrained Problems and Approaches to Solutions

The justication for the use of the Lagrange method is that these
x-values are precisely the solutions to the original constrained
maximization problem we are interested in.

To see this, rewrite the rst two conditions as

f1 (x∗1 , x∗2 ) = −λ∗ g1 (x∗1 , x∗2 )

f2 (x∗1 , x∗2 ) = −λ∗ g2 (x∗1 , x∗2 )

Then taking ratios of the left-hand and right-hand sides respectively

gives
f1 (x∗1 , x∗2 ) g1 (x∗1 , x∗2 )
=
f2 (x∗1 , x∗2 ) g2 (x∗1 , x∗2 )
which is precisely the tangency condition we obtained earlier.

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 77 / 92

Constrained Optimization
Constrained Problems and Approaches to Solutions

Thus eliminating the Lagrange multiplier from the rst two conditions
in this way gives us the tangency condition, and then the third
condition is simply the constraint so that we can solve to obtain the
same solution as before.

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 78 / 92

Constrained Optimization
Constrained Problems and Approaches to Solutions

Thus eliminating the Lagrange multiplier from the rst two conditions
in this way gives us the tangency condition, and then the third
condition is simply the constraint so that we can solve to obtain the
same solution as before.

Example
Solve the constrained maximization problem

max y = x0.25
1 x2
0.75
s.t 100 − 2x1 − 4x2 = 0

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 78 / 92

Constrained Optimization
Constrained Problems and Approaches to Solutions

Thus eliminating the Lagrange multiplier from the rst two conditions
in this way gives us the tangency condition, and then the third
condition is simply the constraint so that we can solve to obtain the
same solution as before.

Example
Solve the constrained maximization problem

max y = x0.25
1 x2
0.75
s.t 100 − 2x1 − 4x2 = 0

Solution
The Lagrange function is

L = x0.25
1 x2
0.75
+ λ(100 − 2x1 − 4x2 )

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 78 / 92

Constrained Optimization
Constrained Problems and Approaches to Solutions

Solution (continued )
and the rst-order conditions are

0.25x−0.75
1 x0.75
2 − 2λ = 0 (8)

−0.25
0.75x0.25
1 x2 − 4λ = 0 (9)

100 − 2x1 − 4x2 = 0 (10)

Solving equations (8) and (9) to eliminate λ gives

x2 = (3/2)x1

and substituting this into equation (10) gives the solution

600 300
x∗1 = , x∗2 = □
48 16
Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 79 / 92
Constrained Optimization
Constrained Problems and Approaches to Solutions

In the discussion so far, we have considered only maximization

problems. The case of minimization problems, however, is treated
similarly.

In Figure we show a strictly convex function f (x1 , x2 ) and a

constraint curve G corresponding to the constraint g(x1 , x2 ) = 0.

The problem is to reach the lowest possible point on the function

while remaining on the constraint curve. The solution is clearly at
point (x∗1 , x∗2 ), a point of tangency.

In fact the level curve diagram looks just like that in the maximization
problem, the dierence being that the value of the function falls as we
move inward toward the point T.

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 80 / 92

Constrained Optimization
Constrained Problems and Approaches to Solutions

Figure

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 81 / 92

Constrained Optimization
Constrained Problems and Approaches to Solutions

Thus, exactly as before, we can show that the optimal point must
satisfy the tangency conditions

f1 (x∗1 , x∗2 ) g1 (x∗1 , x∗2 )

∗ ∗ =
f2 (x1 , x2 ) g2 (x∗1 , x∗2 )

g1 (x∗1 , x∗2 ) = 0

Therefore, we apply the same Lagrange method to nd a solution for

the case of constrained minimization.

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 82 / 92

Constrained Optimization
Constrained Problems and Approaches to Solutions

Thus, exactly as before, we can show that the optimal point must
satisfy the tangency conditions

f1 (x∗1 , x∗2 ) g1 (x∗1 , x∗2 )

∗ ∗ =
f2 (x1 , x2 ) g2 (x∗1 , x∗2 )

g1 (x∗1 , x∗2 ) = 0

Therefore, we apply the same Lagrange method to nd a solution for

the case of constrained minimization.

Example
Solve the following constrained minimization problem:

1/2
min y = x1 + x2 s.t. 1 − x1 − x2 = 0

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 82 / 92

Constrained Optimization
Constrained Problems and Approaches to Solutions

Solution
The Lagrange function is

1/2
L = x1 + x2 + λ(1 − x1 − x2 )

The rst-order conditions are

 
λ −1/2
1− x1 =0
2
1−λ=0
1/2
1 − x1 − x2 = 0

which solve to give

1 1
x∗1 = , x∗2 = , λ∗ = 1 □
4 2
Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 83 / 92
Homogeneous Functions and Euler's theorem
Denition 4.4 (Homogeneous Functions)
The function y = f (x1 , x2 , . . . , xn ) is said to be homogeneous of degree k
if
f (λx1 , λx2 , . . . , λxn ) = λk f (x1 , x2 , . . . , xn )

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 84 / 92

Homogeneous Functions and Euler's theorem
Denition 4.4 (Homogeneous Functions)
The function y = f (x1 , x2 , . . . , xn ) is said to be homogeneous of degree k
if
f (λx1 , λx2 , . . . , λxn ) = λk f (x1 , x2 , . . . , xn )

Example
Given the function f (x, y, w) = x/y + 2w/3x, if we multiply each variable
by j, we get

(jx) 2(jw) x 2w
f (jx, jy, jw) = + = + = f (x, y, w) = j 0 f (x, y, w)
(jy) 3(jx) y 3x

That is, the value of the function will not be aected at all by equal
proportionate changes in all the independent variables; the value of the
function is changed by a multiple of j 0 (= 1). This makes the function f a
homogeneous function of degree zero.

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 84 / 92

Economic Applications
Utility Maximization and Consumer Demand

First-Order Condition
The Lagrangian function of this optimization model is

L = U (x, y) + λ(B − xPx − yPy )

As the rst-order condition, we have the following set of simultaneous

equations:

Lx = Ux − λPx = 0
Ly = Uy − λPy = 0
Lλ = B − xPx − yPy = 0

Since the rst two equations are equivalent to

Ux Uy
= =λ (13)
Px Py

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 85 / 92

Economic Applications
Utility Maximization and Consumer Demand

The rst-order condition calls for the satisfaction of (13), subject to

the budget constraint.

What (13) states is merely the familiar proposition in classical

consumer theory that, in order to maximize utility, consumers must
allocate their budgets so as to equalize the ratio of the marginal utility
to price for every commodity.

If we restate the condition in (13) in the form

Ux Px
= (14)
Uy Py

the rst-order condition can be given an alternative interpretation, in

terms of indierence curves.

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 86 / 92

Economic Applications
Utility Maximization and Consumer Demand

An indierence curve is dened as the locus of the combinations of x

and y that will yield a constant level of U. This means that on an
indierence curve we must nd

dU = Ux dx + Uy dy = 0

with the implication that dy/dx = −Ux /Uy .

Accordingly, if we plot an indierence curve in the xy plane, as in
Figure , its slope, dy/dx, must be equal to the negative of the
marginal-utility ratio Ux /Uy .

Note that Ux /Uy , the negative of the indierence-curve slope, is

called the marginal rate of substitution between the two goods.

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 87 / 92

Constrained Optimization
Constrained Problems and Approaches to Solutions

Figure

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 88 / 92

Economic Applications
Utility Maximization and Consumer Demand

On the other hand, the ratio Px /Py represents the negative of the
slope of the graph of the budget constraint.

The budget constraint, xPx + yPy = B , can be written alternatively as

B Px
y= − x
Py Py

so that, when plotted in the xy plane as in Figure , it emerges as a

straight line with slope −Px /Py (and vertical intercept B/Py ).
In this light, the new version of the rst-order condition(14) plus
the budget constraintdiscloses that, to maximize utility, a consumer
must allocate the budget such that the slope of the budget line is
equal to the slope of some indierence curve.

This condition is met at point E in Figure , where the budget line is

tangent to an indierence curve.

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 89 / 92

Economic Applications
Linear Homogeneity and Production Functions

In the discussion of production functions, wide use is made of

homogeneous functions of the rst degree.

These are often referred to as linearly homogeneous functions.

Let us consider the following production function

Q = f (K, L)

Whether applied at the micro or the macro level, the mathematical

assumption of linear homogeneity would amount to the economic
assumption of constant returns to scale.
This is because linear homogeneity means that raising all inputs
(independent variables) j−fold will always raise the output (value of
the function) exactly j−fold also.

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 90 / 92

Economic Applications
Linear Homogeneity and Production Functions

Example
(a) What is the degree of homogeneity of the Cobb-Douglas production
function

Q = ALα K β
(b) Determine the degree of homogeneity for the following CES
production function

Q = A(αLθ + βK θ )1/θ

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 91 / 92

 ************* End of Chapter Four *************

Ephrem Andargie (AAUSC) Ch 4: Functions of Several Variables October 4, 2023 92 / 92