ECON 3061

Topic II: Calculus of functions of n variables

Diaa Noureldin
[email protected]

Department of Economics
The American University in Cairo

Econ 3061 - Mathematics for Economists II

Outline for Topic II: Calculus of functions of n variables

The material for this topic is Ch. 11 (except Section 11.6) in the course
textbook (Hoy et al., 2011).

I Partial dierentiation.
I Second-order partial derivative.
I The rst-order total dierential.
I Implicit dierentiation.
I Level curves.
I Homogenous and homothetic functions.
I Economic applications.

1 / 61
RECALL: the derivative

Given y = f (x ), which is a function of one variable. The derivative is

dened as
dy f (x + x ) f (x )
f 0 (x ) = lim .
dx x !0 x
dy
The notation dx is used to denote the derivative of y with respect to
(w.r.t.) x.

The second derivative is dened as

d2y
f 00 (x ),
dx 2

and generally the nth derivative is

dny
f (n ) ( x ) .
dx n

2 / 61
RECALL: the rules of dierentiation I

Recall the following rules of dierentiation:

1. f (x ) = c (c is constant), then f 0 (x ) = 0.

2. f (x ) = cx (c is constant), then f 0 (x ) = c.

3. f (x ) = x n , then f 0 (x ) = nx n 1.

4. f (x ) = cx n (c is constant), then f 0 (x ) = ncx n 1.

5. h(x ) = f (x ) g (x ), then h0 (x ) = f 0 (x ) g 0 (x ).

6. h(x ) = f (x )g (x ), then h0 (x ) = f 0 (x )g (x ) + f (x )g 0 (x ).
f (x ) f 0 (x )g (x ) f (x )g 0 (x )
7. h(x ) = , g (x ) 6= 0, then h0 (x ) = .
g (x ) (g (x ))2

3 / 61
RECALL: the rules of dierentiation II
The rules of dierentiation (continued):

8. If y = f (u ) and u = g (x ), then y = f (g (x )) = h(x ) where h is a

composite function. Then we have the chain rule:

h 0 (x ) = f 0 (u )g 0 (x ),

or
dy dy du
= .
dx du dx

Example
Let y = (3x 4)6 , thus u = g (x ) = 3x 4, and y = f (u ) = u 6 . We
have
dy dy du
= = 6u 5 (3) = 18u 5 = 18 (3x 4)5 .
dx du dx

4 / 61
RECALL: the rules of dierentiation III
The rules of dierentiation (continued):

9. If y = expfg (x )g e g (x ) , then

dy
= e g (x ) g 0 ( x ) .
dx

Example
2 dy 2 2
Let y = e 2x 1, then dx = e 2x 1 (4x ) = 4xe 2x 1.

10. If y = log(g (x )), then

dy 1 0
= g (x ).
dx g (x )

Example
dy 1 4x 4x .
Let y = log(2x 2 1), then dx = 2x 2 1
= 2x 2 1

5 / 61
RECALL: the rules of dierentiation IV
The rules of dierentiation (continued):

11. If y = f (x ), where f ( ) is a continuous function then x = f 1 (y ).

Provided that f 0 (x ) 6= 0, we have

dx 1 1
= 0 = .
dy f (x ) dy /dx

Example
dy dx 1
Let y = 3x 2 =) dx = 6x =) dy = 6x . We also have

y 1 /2
x =
3
1 /2
dx 1 y 1 /2 1 1 y 1 /2 1 3x 2
= = =
dy 2 3 3 6 3 6 3
1 /2 1 /2
1 2 1 1 1
= x = = .
6 6 x2 6x

6 / 61
RECALL: the rules of dierentiation V
The rules of dierentiation (continued):

f (x ) 0
12. LHbo pitals rule: If limx !a g (x ) = 0 (i.e. indeterminate), then the
limit can be evaluated as follow:
f (x ) f 0 (x )
lim = lim 0 .
x !a g (x ) x !a g (x )

Example
2 x 2 25
Evaluate limx !5 xx 25 .
5 Since limx !5 x 5 = 00 , then we can use
LHb
o pitals rule:

f (x ) f 0 (x )
lim = lim
x !a g (x ) x !a g 0 (x )
x2 25 2x
lim = lim = 10.
x !5 x 5 x !5 1

7 / 61
RECALL: The total dierential
If f 0 (x0 ) is the derivative of f (x ) at the point x0 , then the total
dierential at the point x0 is
dy = f 0 (x0 )dx.

Example
Let y = x 2 . Let x change from the value 10 to 10.1 (i.e. x = 0.1). We
have that the corresponding y value will change from y = (10)2 = 100
to y = x 2 = (10.1)2 = 102.01. Thus y = 2.01. Using the total
dierential we have

dy = f 0 (x0 )dx
= 20(0.1)
= 2
2.01

noting that f 0 (x ) = 2x and f 0 (x0 = 10) = 2(10) = 20.

You can check that the approximation improves if x is taken to be
smaller and smaller. The opposite is also true. The approximation will be
poorer for a large value of x.
8 / 61
Functions of n variables: Notation
We shall use the following notation:
I x 2 Rn denotes a collection of n points x = (x1 , x2 , ..., xn ).
I y = f (x1 , x2 , ..., xn ) = f (x) denotes a multivariate function f of n
variables.
I Here, Rn is the domain of the function.
I We sometimes use the notation f : Rn ! R, which reads as: the
function f maps from Rn into R.
Example
The output of the rm (y ) depends on two inputs: capital (x1 ) and labor
(x2 ). So we have y f (x) = f (x1 , x2 ), where f : R2 ! R, that is the
function f maps points in R2 (i.e. dierent capital and labor
combinations) into a single value on the real line (R). We can specify
the functional form of f (x) as follows:

y f (x) = Ax1 x2 ,

where A, and are constants. This is known as the Cobb-Douglas

production function.
9 / 61
The partial derivative I
Recall that the derivative
dy f (x + x ) f (x )
f 0 (x ) = lim
dx x !0 x
is interpreted as the rate at which y changes when x changes as we let
the change in x become arbitrarily small.

We can extend this idea to multivariate functions as follows: If

y = f (x1 , x2 ), we can dene the rate at which y changes when x1
changes (keeping x2 constant), and the rate at which y changes when x2
changes (keeping x1 constant). These are called the partial derivatives
of f with respect to its arguments.
Denition
The partial derivatives of f (x1 , x2 ) w.r.t. x1 and x2 , respectively, are
f (x1 , x2 ) y f (x1 + x1 , x2 ) f (x1 , x2 )
f1 (x 1 , x 2 ) = lim
x1 x1 x1 !0 x1
f (x1 , x2 ) y f (x1 , x2 + x2 ) f (x1 , x2 )
f2 (x1 , x2 ) = lim .
x2 x2 x2 !0 x2

10 / 61
The partial derivative II

I The reason for calling these expressions partial derivatives, and for
using the notation "" is that we are changing only one of x1 and x2
at a time, despite the fact that y depends on both of them.
f (x1 ,x2 )
I It is important to note that the partial derivative will in
x 1
f (x1 ,x2 )
general depend on x2 , and the partial derivative x 2 will in
general depend on x1 !
I For instance, the rate of change of output w.r.t. to the labor input
will depend on the level of capital input.
I Similiarly, the rate of change of output w.r.t. to the capital input
will depend on the amount of labor employed in production.

11 / 61
The partial derivative: General denition

Denition
The partial derivative of a function y = f (x1 , x2 , ..., xn ) = f (x) w.r.t.
the variable xi is:

f f (x1 , ..., xi + xi , ..., xn ) f (x1 , ..., xi , ..., xn )

fi ( x ) = lim .
xi x i !0 xi

We can also use alternative notation for the general denition of the
partial derivative:

f f (xi + xi jx i ) f (x)
f i ( x ) = f 0 ( xi j x i ) = lim ,
xi x i !0 xi

where the notation f 0 (xi jx i ) denotes the derivative of f (x) w.r.t. to its
i th argument, holding all other xs (apart from xi ) constant.

12 / 61
The partial derivative: Examples I
Example
Let y = f (x1 , x2 ) = x13 + 5x2 . In this example, we have the arguments of
f being additively separable. We have

f
f1 = 3x12
x1
f
f2 = 5.
x2

Example
Let y = f (x1 , x2 ) = x12 x2 . We have

f
f1 = 2x1 x2
x1
f
f2 = x12 .
x2

13 / 61
The partial derivative: Examples II

Example
Let y = f (x1 , x2 ) = 10x1 x22 5x13 + 4x2 + x1 x2 . We have

f
f1 = 10x22 15x12 + x2
x1
f
f2 = 20x1 x2 + 4 + x1 .
x2

You should note that in these examples, the partial derivative will in
general depend on all the arguments of f .

14 / 61
The demand function I

We have that the demand for product a (QaD ) will depend on the price of
a (Pa ), the price of a related good b (Pb ) as well as income (I):

QaD = f (Pa , Pb , I).

We also have the following expected signs for the partial derivatives:
I
Q aD
P a fP a < 0 (law of demand).
I
Q aD
P b fP b > 0 (goods a and b are substitutes).
I
Q aD
P b fP b < 0 (goods a and b are complements).
I
Q aD
I fI > 0 (good a is a normal good).
I
Q aD
I fI < 0 (good a is an inferior good).

15 / 61
The demand function II

Example
Let
QaD = f (Pa , Pb , I) = 50Pa + 10Pb 2I.
Then we have
I
Q aD
P a fP a = 50 (downward sloping demand curve).
I
Q aD
P b fP b = 10 (goods a and b are substitutes).
I
Q aD
I fI = 2 (good a is an inferior good).

16 / 61
The utility function I

Let the utility function be given by

U (x1 , x2 ) = x1 x21
.

We have the marginal utility functions ( xU1 and U

x2 ).

I U 1 1
x 1 U1 = x1 x2 , which is positive for > 0.

I U U2 = x1 (1 )x2 , which is positive for

x 2
1 > 0 () < 1.
I Therefore, for U to be a feasible utility function, we need 0 < < 1.

17 / 61
The utility function II

For 0 < < 1, the marginal utility functions satsify the law of
diminishing marginal utility. This is obvious since

U 1 1 x21
= x1 x2 =
x1 x11

U (1 )x1
= x1 (1 )x2
= .
x2 x
2

I U U
We have x 1 > 0, however x 1 # as x1 ".
I U U
We have x 2 > 0, however x 2 # as x2 ".

18 / 61
The production function
Let Q = f (K , L) be a production function for a rm, where Q, K and L
denote output, capital and labor, respectively.
Example
Let Q = 3L1 /2 K 1 /2 . We have
Q 3 1 /2 1 /2
QL = L K
L 2
Q 3
QK = L1 /2 K 1 /2 .
K 2

The quantity QL is known as the marginal product of labor (MPL).

Similarly, Q
K is known as the marginal product of capital (MPK).

The concept of the production function can also be applied on the macro
level, where Q denotes aggregate economic output (or GDP), L denotes
the labor force and K denotes the capital stock (or cumulative
investments).

19 / 61
The Cobb-Douglas production function
The Cobb-Douglas production function is given by

Q = AL K .

I A > 0, 0 < , < 1 are technology parameters.

I The marginal product functions are

Q 1 AK
= AL K =
L L1
Q 1 AL
= AL K = 1 .
K K
I Note that both marginal product functions satisfy the law of
diminshing marginal productivity.
I We have QL > 0; however QL # as L ".
I Q Q
Also K > 0; however K # as K ".

20 / 61
The Cobb-Douglas production function with 3 inputs
With three factors of production, the Cobb-Douglas production function
is given by

Q = Ax1 x2 x3 , A > 0, 0 < , , < 1.

I The marginal product functions are

Q 1 x2 x3
= Ax1 x2 x3 = A
x1 x11

Q 1 x1 x3
= Ax1 x2 x3 = A 1
x2 x2

Q 1 x1 x2
= Ax1 x2 x3 = A 1
.
x3 x3

I All marginal products satisfy the law of diminshing marginal

productivity.
I The marginal product of each input is decreasing in that input;
however it is increasing in the other two inputs.
21 / 61
The Cobb-Douglas production function: Example
Let
Q = 10L1 /3 K 1 /2 .

I Q = 10 L 2 /3 K 1 /2 = 10 K 1/2 .
L 3 3 L 2/3
1/3
I Q

= 1
5L K / 3 1 / 2 L
= 5 K 1/2 .
K
Suppose the total labor force is 25 million workers, and the capital stock
(measured in monetary terms) is LE 15 billion. Then we have

Q 10 K 1 /2 10 (15)1 /2
= = = 1.51
L 3 L2 /3 3 (25)2 /3
Q L1 /3 (25)1 /3
= 5 = 5 = 3.77.
K K 1 /2 (15)1 /2
The gures are interpreted as follows:
I As labor increases by 1 unit (i.e. 1 million workers), Q increases by
LE 1.51 billion.
I As capital increases by 1 unit (i.e. LE 1 billion), Q increases by
LE 3.77 billion.
22 / 61
In-class exercises I

Find the derivatives of the following functions:

2 +3x +2
1. y = e 2x .

2
2. y = e x log x 2 .

3. y = e log (x ) .

23 / 61
In-class exercises II
Find the partial derivatives of the following functions:

2x23
y = x12 e 3x2 +x1 x3 + .
x1

24 / 61
In-class exercises III

1. Find the marginal product functions for the following Cobb-Douglas

production function:

y = 10x11 /2 x21 /3 x31 /4 .

2. Find the marginal product functions for the following CES (constant
elasticity of substitution) production function:
2
1 /2 1 /2
y = 12 0.4x1 + 0.6x2 .

25 / 61
The constant elasticity of substitution (CES) production
function I
The CES production function with 2 inputs is given by:
1 /r
y = A x1 r + (1 ) x2 r
,

where A > 0, 0 < < 1 and r > 1.

The marginal products are

y 1 (1/r ) 1
= A x1 r + (1 ) x2 r
r x1 r 1
x1 r
A (1/r ) 1
= x1 r + (1 ) x2 r
.
x1r +1
y 1 (1/r ) 1
= A x1 r + (1 ) x2 r r (1 ) x2 r 1
x2 r
A (1/r ) 1
= (1 ) r +1 x1 r + (1 ) x2 r .
x2

26 / 61
The CES (constant elasticity of substitution) production
function II
y Ar
The marginal products can be written in a simpler form. Multiply x1 by Ar ,
we have
y Ar A (1/r ) 1
= x1 r + (1 ) x2 r
x1 Ar x1r +1
(1/r ) 1
Ar +1 x1 r + (1 ) x2 r
= .
Ar x1r +1
Also note that
1/r r +1
y r +1 = A x1 r + (1 ) x2 r

(1/r ) 1
= Ar +1 x1 r + (1 ) x2 r
.
Thus we have
r +1
y y r +1 y
= r r +1 = r .
x1 A x1 A x1
And Similarly, we have
r +1
y (1 ) y r +1 (1 ) y
= = .
x2 Ar x2r +1 Ar x2
27 / 61
The CES (constant elasticity of substitution) production
function III

The marginal products are

r +1
y y
=
x1 Ar x1
r +1
y (1 ) y
= .
x2 Ar x2

Note that we have the condition that r > 1, which implies that
r +1 > 0.

This in turn implies

I y
Diminishing marginal productivity of input 1: x 1 # as x1 ".
I y
Diminishing marginal productivity of input 2: x 2 # as x2 ".

28 / 61
Second-order partial derivatives
I Now that we have dened the (rst-order) partial derivative, we now
move on to dene the second-order partial derivatives.
I Recall the denition of the rst-order partial derivatives of f (x1 , x2 ):
f (x1 , x2 ) y f (x1 + x1 , x2 ) f (x1 , x2 )
f1 (x 1 , x 2 ) = lim
x1 x1 x1 !0 x1
f (x1 , x2 ) y f (x1 , x2 + x2 ) f (x1 , x2 )
f2 (x1 , x2 ) = lim .
x2 x2 x2 !0 x2

I A function of n variables, y = f (x1 , x2 , ..., xn ), will have n rst-order

partial derivatives f1 , f2 , ..., fn .
I The second-order partial derivatives are given by

fi (x1 , x2 , ..., xn )
fij = , i , j = 1, 2, ..., n.
xj

I fij is found by rst dierentiating the function f (x) w.r.t. xi , and then
dierentiation the result w.r.t. xj .

29 / 61
Second-order partial derivatives (continued)
It is clear that for a function of n variables, there will be n2 second-order
partial derivatives.

Example
For y = f (x1 , x2 ) we have 4 second-order partial derivatives:

f11 f12
f21 f22

Example
For y = f (x1 , x2 , x3 ) we have 9 second-order partial derivatives:

f11 f12 f13

f21 f22 f23
f31 f32 f33

This makes it clear that we can neatly present the second-order partial
derivatives using matrices.
30 / 61
The gradient vector

First let us dene the gradient vector, denoted by 5f , which collects all
the rst-order partial derivatives of the n-variable function f (x).

Denition
The gradient vector is 2 3
f1
6 f2 7
6 7
5f = 6 .. 7.
4 . 5
fn

The reason is it called the gradient vector is that each element fi

indicates the rate of change in f (x) w.r.t. xi . This analogous to signs
indicating the grades or steepness that we nd when driving a car
through hills or mountains.

31 / 61
The gradient vector: Examples

Example
Find the gradient vector for f (x) = 2x12 + 3x2 + x34 + 1. We have
2 3 2 3
f1 4x1
5 f = 4 f2 5 = 4 3 5 .
f3 4x33

Note that this function is additively separable.

Example
Find the gradient vector for f (x) = x12 x23 x34 . We have
2 3 2 3
f1 2x1 x23 x34
5f = 4 f2 5 = 4 3x12 x22 x34 5 .
f3 4x12 x23 x33

32 / 61
The Hessian matrix

The Hessian matrix, denoted by 52 f , collects the second-order partial

derivatives.

Example
For y = f (x1 , x2 ), we have n2 = 22 = 4 second-order partial derivatives:

f1 (x1 , x2 ) f1 (x1 , x2 )
f11 , f12
x1 x2
f2 (x1 , x2 ) f2 (x1 , x2 )
f21 , f22 .
x1 x2
The Hessian matrix is given by

f11 f12
52 f = .
f21 f22

33 / 61
The Hessian matrix (continued)

For n = 3, the Hessian matrix is given by

2 3
f11 f12 f13
52 f = 4 f21 f22 f23 5 .
f31 f32 f33

And generally for any n, the Hessian is given by

2 3
f11 f12 f1n
6 f21 f22 f2n 7
6 7
52 f = 6 .. .. . . .. 7 .
4 . . . . 5
fn1 fn2 fnn

34 / 61
The Hessian matrix: Example I
Find the Hessian matrix for f (x) = x12 x23 .

Solution
First, the gradient vector (i.e. vector of rst-order partial derivatives) is

f1 2x1 x23
5f = = .
f2 3x12 x22

We now compute

f1 f1
f11 = 2x23 , f12 = 6x1 x22 ,
x1 x2
f2 f2
f21 = 6x1 x22 , f22 = 6x12 x2 ,
x1 x2
and we have
f11 f12 2x23 6x1 x22
52 f = = .
f21 f22 6x1 x22 6x12 x2

35 / 61
The Hessian matrix: Example II
Find the Hessian matrix for f (x) = x12 x23 x34 .

Solution
The gradient vector (i.e. vector of rst-order partial derivatives) is
2 3 2 3
f1 2x1 x23 x34
5f = 4 f2 5 = 4 3x12 x22 x34 5 .
f3 4x12 x23 x33

We now compute
f1 f1 f1
f11 = 2x23 x34 , f12 = 6x1 x22 x34 , f13 = 8x1 x23 x33 ,
x1 x2 x3
f2 f2 f2
f21 = 6x1 x22 x34 , f22 = 6x12 x2 x34 , f23 = 12x12 x22 x33 ,
x1 x2 x3
f3 f3 f3
f31 = 8x1 x23 x33 , f32 = 12x12 x22 x33 , f33 = 12x12 x23 x32 ,
x1 x2 x3
and we have
2 3 2 3
f11 f12 f13 2x23 x34 6x1 x22 x34 8x1 x23 x33
52 f = 4 f21 f22 f23 5 4
= 6x1 x22 x34 6x12 x2 x34 12x12 x22 x33 5 .
f31 f32 f33 8x1 x23 x33 12x12 x22 x33 12x12 x23 x32 36 / 61
Symmetry of the Hessian matrix

We notice from the previous examples that the Hessian matrix is

symmetric, i.e. fij = fji 8i, j.

Does this hold in all cases? The answer is YES.

Theorem
( Youngs Theorem) For a function y = f (x1 , x2 , ..., xn ) with
continuous rst- and second-order partial derivatives, the order of
dierentiation in computing the cross-partials does not matter. That is
fij = fji 8i, j = 1, 2, ..., n.

37 / 61
The Hessian matrix for additively separable functions I

I Recall the additively separable function f (x) = 2x12 + 3x2 + x34 + 1

(see above). We have
2 3 2 3
f1 4x1
5 f = 4 f2 5 = 4 3 5 .
f3 4x33

such that f1 is a function of only x1 , f2 is a function of only x2

(constant function in this case), and f3 is a function of only x3 .

I It is obvious that this feature holds in general for any additively

separable function.

I This immediatley implies that all the cross-partials will be zeros, and
the Hessian matrix will be diagonal in this case.

38 / 61
The Hessian matrix for additively separable functions II

For the additively separable function f (x) = 2x12 + 3x2 + x34 + 1, it is

straightforward to show that the Hessian matrix is given by
2 3 2 3
f11 f12 f13 4 0 0
52 f = 4 f21 f22 f23 5 = 4 0 0 0 5.
f31 f32 f33 0 0 12x32

Verify this as an exercise.

39 / 61
RECALL: The total dierential
If f 0 (a) is the derivative of f (x ) at the point x = a, then the total
dierential at the point x = a is
dy = f 0 (a)dx.
The total dierential is used to approximate the change in y (y ) for a
given change in x (x) within a neighbourhood of the point x = a. That
is
y = f (a + x ) f (a)
can be approximated by
dy = f 0 (a)dx.

40 / 61
The total dierential for a function of 2 variables
For y = f (x) = f (x1 , x2 ), where f : R2 ! R, the total dierential is
f f
dy = dx1 + dx2 = f1 dx1 + f2 dx2 .
x1 x2
The interpretation here is as follows:
I At a specic point where x1 = a1 and x2 = a2 , i.e. y = f (a1 , a2 ), we
could change only x1 and get the result
y = f (a1 + x1 , a2 ) f (a1 , a2 ),
which can be approximated by
dy = f1 (a1 , a2 )dx1 .

I Similarly, we could change only x2 and get the result

y = f (a1 , a2 + x2 ) f (a1 , a2 ),
which can be approximated by
dy = f2 (a1 , a2 )dx2 .

I Allowing for both x1 and x2 to change leads to the above denition.

41 / 61
The total dierential: A geometric view
Plot of y = f(x ,x ) = 1 - x 2 - x 2
1 2 1 2

10

-10

-20
y

-30

-40

-50
5
0 5
0
-5 -5
x x
2 1

42 / 61
The total dierential for a function of n variables

For y = f (x) = f (x1 , ..., xn ), where f : Rn ! R, the total dierential is

f f f
dy = dx1 + dx2 + + dxn
x1 x2 xn
= f1 dx1 + f2 dx2 + + fn dxn
n
= fi dxi .
i =1

In this multivariate case, the total dierential is used to approximate

changes in y given simultaneous changes in the inputs xi , i = 1, 2, ..., n.

43 / 61
The total dierential: Example

Let
1
y = f (x) = 2x12 + x1 x2 3x32 .
2
Find the total dierential for f (x).

Solution
The total dierential is given by

dy = f1 dx1 + f2 dx2 + f3 dx3

1 1
= 4x1 + x2 dx1 + x1 dx2 (6x3 ) dx3 .
2 2

44 / 61
Implicit dierentiation I
When we have y = f (x ), we have an explicit function.
Example
dy
The function y = 3x 2 + 4, where dx = 6x.

Sometimes we have F (x, y ) = c, c is a constant, which is called an

implicit function.
Example
For
x 2 y 3 + 3xy 2 + y = 22
which has the form F (x, y ) = c. This is an example where y is an
implicit function of x, since changes in x induce changes in y for the
equation to hold.

We need to nd a rule to be able to nd dy dx when we are dealing with

implicit functions. This is called implicit dierentiation.

45 / 61
Implicit dierentiation II
To deal with this problem, we totally dierentiate the implicit function
F (x, y ) = c as follows:
Fx dx + Fy dy = 0
Fy dy = Fx dx
which in turn implies
dy Fx
= , Fy 6= 0.
dx Fy

Example
For
x 2 y 3 + 3xy 2 + y = 22,
we have
dy Fx 2xy 3 + 3y 2
= = .
dx Fy (3x 2 y 2 + 6xy + 1)

This is an example of the application of the Implicit Function Theorem

which is stated on the following slide.
46 / 61
The Implicit Function Theorem
The Implicit Function Theorem can be stated as follows: Given
F (x1 , x2 , ..., xn , y ) = c, and assuming Fy 6= 0, then the partial
derivatives are given by
y Fxi
= .
xi Fy

Example
Compute the partial derivatives of

F (x1 , x2 , y ) = 3x1 x2 + x2 y 2 + x12 x2 y = 10.

Here we can use the Implicit Function Theorem to compute the partial
derivatives. We have
y Fx1 (3x2 + 2x1 x2 y )
= =
x1 Fy 2x2 y + x12 x2
y Fx2 3x1 + y 2 + x12 y
= = .
x2 Fy 2x2 y + x12 x2

47 / 61
Level Curves
Suppose we have an explicit function y = f (x1 , x2 ). Then
y = f (x1 , x2 ) = c, c is a constant,
is the level curve of y .

Using the total dierential, we have

dy = f1 dx1 + f2 dx2 = 0
f2 dx2 = f1 dx1
dx2 f1
= .
dx1 f2

Example
Let y = f (x1 , x2 ) = 2x1 + 3x2 . Then the level curve for this function is

y = 2x1 + 3x2 = c,

where c can take dierent values. Also we have

dx2 f1 2
= = .
dx1 f2 3
48 / 61
Level curves: The Indierence curve
Let U (x1 , x2 ) be the utility function from consumption of goods x1 and
x2 . The level curves are given by U (x1 , x2 ) = c.

Using the total dierential, we have

U U
dU = dx1 + dx2 = 0.
x1 x2
U U
Dene MU1 x 1 and MU2 x 2 . Then we have

MU1 dx1 + MU2 dx2 = 0

MU2 dx2 = MU1 dx1
dx2 MU1
=
dx1 MU2
dx2 MU1
= = MRS
| {z }
dx1 MU2
| {z } Marg. rate of substitution
Slope of IC

49 / 61
Level curves: The Indierence curve (example)
Let U (x1 , x2 ) = ln(x1 ) + (1 ) ln(x2 ), 0 < < 1. Below is the plot
of U (x1 , x2 ) for = 0.5.
Plot of U(x ,x ) = (1/2)log(x ) + (1/2)log(x )
1 2 1 2

2.5

1.5
U(x ,x )
2
1

0.5

0
10

5 10
8
6
4
2
x 0 0
2 x
1

50 / 61
Level curves: The Indierence curve (example, continued)
The level curves (indierence curves) are given by
U (x1 , x2 ) = ln(x1 ) + (1 ) ln(x2 ) = c ( = 0.5):
Level curves of U(x ,x ) = (1/2)log(x ) + (1/2)log(x )
1 2 1 2
10

6
2
x

1
1 2 3 4 5 6 7 8 9 10
x
1

51 / 61
Level curves: The Indierence curve (numerical example)

For U (x1 , x2 ) = 14 ln(x1 ) + 34 ln(x2 ), nd the marginal rate of

susbtitution between x1 and x2 .

Solution
We rst compute the marginal utilities:

U 1 U 3
MU1 = = , MU2 = = ,
x1 4x1 x2 4x2
and the marginal rate of substitution is given by
1
dx2 MU1 4x1 x2
MRS = = = 3
= .
dx1 MU2 4x2
3x1

52 / 61
Level curves: The Isoquant
Let Q = f (L, K ) be the production function using labor (L) and capital
(K ) inputs. The level curves are given by Q = f (L, K ) = c.

Using the total dierential, we have

f f
dQ = dL + dK = 0.
L K
f f
Dene MPL L and MPK K . Then we have

MPL dL + MPK dK = 0
MPK dK = MPL dL
dK MPL
=
dL MPK
dK MPL
= = MRTS
| {z }
dL MPK
| {z } Marg. rate of tech. substitution
Slope of Isoquant

53 / 61
Level curves: The Isoquant (example)
Let Q = f (L, K ) = L K (1 ) , 0 < < 1. Below is the plot of
Q = f (L, K ) for = 13 .
1/3 2/3
Plot of Q(L,K) = L K

120

100

80

60
Q

40

20

0
10
10
8
5 6
4
2
K 0 0
L

54 / 61
Level curves: The Isoquant (example, continued)
The level curves (isoquants) are given by Q = f (L, K ) = L K (1 ) =c
( = 13 ):
1/3 2/3
Level curves of Q(L,K) = L K
10

5
K

0
0 2 4 6 8 10
L

55 / 61
Level curves: The Isoquant (numerical example)

For Q = f (L, K ) = L1 /3 K 2 /3 , nd the marginal rate of technical

susbtitution between L and K .

Solution
We rst compute the marginal productivities:
2 /3 1 /3
f 1 K f 2 L
MPL = = , MPK = = ,
L 3 L K 3 K

and the marginal rate of technical substitution is given by

2 /3
1 K
dK MPL 3 L K
MRTS = = = 1 /3
= .
dL MPK 2 L 2L
3 K

56 / 61
Homogenous functions I
Denition
A multivariate function y = f (x1 , x2 , ..., xn ) is homogenous of degree
k, i.e. H k , if and only if:

f (x1 , x2 , ..., xn ) = k f (x1 , x2 , ..., xn ).

Example
Let y = f (x1 , x2 ) = 3x1 x2 . We have

f (x1 , x2 ) = 3 (x1 ) (x2 ) = 2 3x1 x2 = 2 f (x1 , x2 ),

therefore it is homogenous of degree 2.

Example
x1
Let y = f (x1 , x2 ) = 3x22
. We have

(x1 ) x
f (x1 , x2 ) = 2
= 2 12 = 1
f (x1 , x2 ),
3 (x2 ) 3x2

therefore it is homogenous of degree 1. 57 / 61

Homogenous functions II: Cobb-Douglas

Example
Let Q = f (L, K ) = L K , 0 < , < 1. We have

f (L, K ) = (L) (K ) = + L K = + f (L, K ),

therefore it is homogenous of degree + .

+ Homog. Returns to Interpretation

of degree Scale (RtS)
1 1 (or H 1 ) Const.RtS Inputs " by , output " by
>1 > 1 (or H >1 ) Inc. RtS Inputs " by , output " by more than
<1 < 1 (or H <1 ) Dec. RtS Inputs " by , output " by less than

58 / 61
Eulers Theorem
If a multivariate function y = f (x1 , x2 , ..., xn ) is homogenous of degree k
(i.e. H k ), then
n
f i xi = ky .
i =1

Example
Cobb-Douglas production function: From above we have that
Q = f (L, K ) = L K is H + . Then

fL L + fK K = ( + ) Q.

Thus Eulers Theorem relates the marginal productivities, inputs, output

and returns to scale in a single equation.

You are required to derive the result in the example above as an

exercise.

59 / 61
Homothetic functions

Denition
If a multivariate function y = f (x1 , x2 , ..., xn ) is homogenous of degree k
(i.e. H k ), then g (y ) is homothetic if g 0 (y ) > 0.

As implied by the denition, homothetic functions are monotonic

transformations of homogenous functions. A monotonic tranformation is
any function g ( ) where g 0 ( ) > 0.

Examples of monotonic transformations

y
g (y ) y + 2 2y 2 y 2 ln (y ) e y
1 1
g 0 (y ) 1 2 2 2y y ey

You will study the properties of homothetic functions in more detail in

intermediate and advanced microeconomics.

60 / 61
Homothetic functions: Example

Let y = f (x1 , x2 ) = x1 x2 . We have that

f (x1 , x2 ) = (x1 ) (x2 ) = 2 x1 x2 = 2 y .

Therefore this is a function that is H 2 .

Example
The function g (y ) = y + 10 = x1 x2 + 10 is a homothetic function since
it is a monotonic transformation of a homogenous function, where
g 0 (y ) > 0.

Example
The function g (y ) = ln (y ) = ln (x1 x2 ) is a homothetic function since it
is a monotonic transformation of a homogenous function, where
g 0 (y ) > 0.

61 / 61