Mathematics I & II

Preliminaries

Dr. Clemens Buchen

clemens.buchen@ebs.edu

Fall Term 2018

Spring Term 2019
Mathematics I & II

I These courses provide the basis for an understanding of

economic models in many applications. Further courses:
Microeconomics (2nd semester), Macroeconomics,
Economics of the Firm (3rd semester), economics courses in
the 5th and 6th semesters, Statistics, Econometrics, Finance,
Operations, ...
I Ultimate goal: get a feeling for the economic way of thinking
and acquire the ability to formalize ideas.
I A lot of work: in total 90 hours, including lectures and tutorials,
over the course of the semester. The lecture gives an
overview of the material, the tutorials provide some exercises.
However, the bigger part of your work takes place outside the
classroom!
I We will proceed fast.
If there is something you don’t understand, this is a hierarchy of
steps you could take:
1. Read the book
2. Sleep on it and read the book again
3. Ask your friend
4. Ask your tutor or ask me
I Talk to me after the lecture
I Come to my office (you can step by unannounced, but
sometimes I might not be there or will have to make an
appointment if I am busy)
I Write an email (clemens.buchen@ebs.edu). If you do write an
email, please make my life easy by stating as exactly as you
can how you would solve a problem.
I Make an attempt to study and understand the proofs.
Although it is time-consuming and frustrating at times,
eventually the rewards are great. Don’t “cram for the test”.
I If you are stuck with something, revisit it later. Don’t look at
the solution right away
I Buy the book: Sydsaeter/Hammond, 2016, Essential
Mathematics for Economics Analysis, 5th ed., Prentice Hall.
I Lectures, tutorials and online platform MyMathLab
I All material (lecture slides, tutorial assignments with solutions)
is availabe on the course site on CampusNet.
MyMathLab

The online tool should be used in three ways:

1. Do the “homework” (no credit)
2. Independent study (no credit)
3. Take tests (credit)
I Dates will be confirmed
On CampusNet you find instructions for enrollment. You need to
first do the Orientation homework to get started.
Mathematics I
Module Quantitative Methods I

Lecture 1 and 2 Differentiation (Ch. 6)

Lecture 3 and 4 ... and some applications (Ch. 7)
Lecture 5 and 6 Functions of one variable and optimization (Ch. 8)
Lecture 7 and 8 Integration (Ch. 9)
Lecture 9 and 10 Functions of two and more variables (Ch. 11)
Lecture 11 and 12 Comparative Statics (Ch. 12)
Lecture 13 and 14 Multivariate Optimization (Ch. 13)
Exam
Mathematics II
Module Quantitative Methods II, Spring Term 2019

Lecture 1 and 2 Constrained Optimization (Ch. 14)

Lecture 3 and 4 Constrained Optimization (Ch. 14)
Lecture 5 and 6 Linear Algebra (Ch. 15)
Lecture 7 and 8 Linear Algebra (Ch. 16)
Lecture 9 and 10 Game Theory
Lecture 11 and 12 Game Theory
Lecture 13 and 14 Game Theory
Exam
Examination

Type Credits
Mathematics I Online I 15
Mathematics I Online II 15
Mathematics I Written 60
90
The exam schedule for Mathematics II is the same.
Differentiation
Mathematics I
Lectures 1-2

Dr. Clemens Buchen

1 / 26
Plan of the lecture

I Applications
I Formal definition of derivatives
I Rules of differentiation
I Increasing and decreasing functions
I Convexity and concavity
I Exponential and logarithmic functions
Based on chapter 6

2 / 26
Applications of Differentiation

I Optimization problems; Differentiation helps us locate maxima and

minima.
I Characterization of maximum/minimum points; Differentiation helps us
form an economic intuition about what determines maxima and minima.
In an abstract way we attempt to understand how to choose the optimum.
I “Comparative statics”; Differentiation helps us understand how
equilibria change if some parameters of the environments change.
Example: what happens to an equilibrium in a market if some demand
parameters change?

3 / 26
Rates of Change
Example 1
Consider a square with lengths x. How does the area of the square change if
the length of the sides increases?

Δx

x x

x x Δx

4 / 26
 Slopes
Slope of Curves
Even though the
Consider in economics
graph of awefunction
are usually
f :Rinterested in the is
7→ R. What derivative
measuredas aby
rate
theof change, we
begin this chapter with
steepness of its graph? a geometrical motivation for the concept. When we study the graph
of a function, we would like to have a precise measure of the steepness of the graph at a point.
y
L

P
f (a)
y ! f (x)

a x
′
Figure 1 f (a) = 1/2
P has coordinates (a, f (a)). The slope of the tangent to the graph at P is called
the derivative of f at point a, and we denote this number by f 0 (a) (read as “f
prime a”). In general, we have

f 0 (a) = the slope of the tangent to the curve y = f (x) at the point (a, f (a))
onomic Analysis EME3_C06.TEX, 25 February 2008, 13:25 P
5 / 26
 6.2 The Derivative. Tangents
The previous section gave a rather vague definition of the tangent to
Geometric interpretation
we said is that it is a straight line which just touches the curve at tha
more formal definition of the same concept.
y y
T T

P P

Figure 1 Figure 2

The straight line through P and Q is called a secant. Keeping P fixed and
Thethe
letting Q rotate along geometrical ideathe
curve yields behind the definition
limiting is easy
straight line PTtowhich
understand.
is C
curve in the xy-plane
called the tangent to the curve at P. (see Fig. 1). Take another point Q on the cur
line through P and Q is called a secant. If we keep P fixed, but let Q
toward P , then the secant will rotate around P , as indicated in Fig.
6 / 26 2
line through P and Q is called a secant. If we keep P fixed, but let Q move along the curve
toward P , then the secant will rotate around P , as indicated in Fig. 2. The limiting straight
line P T toward which the secant tends is called the tangent (line) to the curve at P .
Suppose that P is a point on the graph of the function f . We shall see how the preceding
First consider
considerations the slope
enable of the
us to find the secant
slope ofPQ:
the tangent at P . This is shown in Fig. 3.
y
T

Q ! (a " h, f (a " h))

f
f (a " h) # f (a)
h
P ! (a, f (a))

Figure 3

! "
Point P in Fig. 3 has the coordinates a, f (a) . Point Q lies close to P and is also on the
graph of f . Suppose that the x-coordinate of Q is a + h, where h is a small number ̸ = 0.
Then the x-coordinate of Q is not a (because Q ̸ = P ), but a number close to a. Because Q
7 / 26
 I Analytically, the slope of the secant PQ is measured by mPQ :

f (a + h) − f (a)
mPQ =
h
I In order to get the slope of the tangent PT, we let h converge to 0:

Definition of derivative
f (a + h) − f (a)
f 0 (a) = lim
h→0 h
To summarize, we can interpret the derivative f 0 (a) as the instantaneous rate
of change of a functional value f at the point a.

8 / 26
 Figure 4 provides a geometric interpretation of (∗). In Fig. 5, we hav
to the curve y = x 2 corresponding to a = 1/2 and a = −1. At
f (a) = (1/2)2 = 1/4 and f ′ (1/2) = 1. According to (2), the equat
y − 1/4 = 1 · (x − 1/2) or y = x − 1/4. (Show that the other tangen
Example 2
the equation y = −2x − 1.) Note that the formula f ′ (a) = 2a shows
2
a < 0,f (x)
Consider the function f ′x(a) > 0 when a > 0. Does this agree with the graph?
and=

y f (x) # x 2 y

Q 4

(a ! h) 2 " a 2 # 2ah ! h 2 2

1
P h

a a!h x "2 "1

"1

Figure 4 f (x) = x 2 Figure 5 f (x) =

If f is a relatively simple function, we can find f ′ (a) by using the fol

9 / 26
Example 3
Suppose we have a cost function C(x). Using the definition of a derivative we
can write an expression for the marginal costs of this firm

C(x + h) − C(x)
C0 (x) = lim
h→0 h
C(x + h) − C(x)
≈ for small h
h
Denoting C0 with MC, we have:

MC(x) ≈ C(x + 1) − C(x) for h = 1.

Intuition
The marginal costs approximately capture the variation in costs if we raise the
quantity produced by 1 unit.

10 / 26
Overview of differentiation rules (1/2)

[Power functions] For any real number a and f (x) = xa :

f 0 (x) = axa−1

[Constants] For any real number A and f (x) = A:

f 0 (x) = 0

[Sums] For F(x) = f (x) + g(x):

F 0 (x) = f 0 (x) + g0 (x)

[Products] For F(x) = f (x) · g(x):

F 0 (x) = f 0 (x)g(x) + f (x)g0 (x)
Overview of differentiation rules (2/2)
f (x)
[Quotients] For F(x) = g(x) :

f 0 (x)g(x) − f (x)g0 (x)

F 0 (x) =
[g(x)]2

[Composite functions] For F(x) = f (g(x)), where f (y) and y = g(x)

F 0 (x) = f 0 (g(x)) · g0 (x)

[Exponential functions] For f (x) = ax :

f 0 (x) = ax · ln a

[Natural logarithm] For f (x) = ln h(x), where h > 0

h0 (x)
f 0 (x) =
h(x)
A useful trick1

Recall the definition of a derivative, now more generally written with a

generic x:
f (x + h) − f (x)
f 0 (x) = lim
h→0 h
Suppose we do not take the limit:

f (x + h) − f (x)
f 0 (x) ≈ ⇐⇒
h
f (x + h) ≈ f (x) + f 0 (x)h
Using this trick, we can prove the product rule and the chain rule.

1 This is called a linear approximation and we will return to this in the next lecture.
13 / 26
Leibniz Notation

Leibniz used a different, yet very intuitive notation for derivatives which is
often used in economics.
I Suppose y = f (x), then Leibniz notation is

df dy
f 0 (x) = or f 0 (x) = (x)
dx dx
Intuition:
f (x + h) − f (x) = dy and h = dx
I The chain rule using Leibniz notation:

df dy
(f (g(x)))0 = where y = g(x)
dy dx

14 / 26
 If f (x2 ) ≤ f (x1 ) whenever x2 > x1 , then f is decreasing in I
If f (x2 ) < f (xand
Increasing 1 ) whenever then f is strictly decreasing in I
x2 > x1 ,functions
decreasing

y y y y

x x x x
Increasing Strictly Decreasing Strictly
increasing decreasing
Figure 1

I f 0 (x) ≥ 0 for all x in the interval I ⇐⇒ f is increasing in I

Figure 1 illustrates these definitions. Note that we allow an increasing (or decreasing) fun
tion f 0 (x)
I to have≤sections
0 for allwhere
x in the interval
the graph I ⇐⇒ f is
is horizontal. decreasing
This in I agree with comm
does not quite
0
I f (x) > I ⇐⇒
language. Few people
0 for all xwould
in thesay that their
interval salaryf increases
is strictlywhen it stays constant!
increasing in I To fi
out
I on 0 which intervals a function is (strictly) increasing or (strictly)
f (x) < 0 for all x in the interval I ⇐⇒ f is strictly decreasing in I decreasing using t
definitions,
I f 0 (x) =we0 have to xconsider
for all the sign of
in the interval I f⇐⇒
(x2 )−f 1 ) whenever
f is(xconstant in I x2 > x1 . This is usua
quite difficult to do directly by checking the values of f (x) at different points x. In fact, w
15 / 26
Example 4
A plot of the function f = x2 and its derivative f 0 = 2x
25

20

15

10

-4 -2 2 4

-5

-10

16 / 26
Higher-Order Derivatives

The second derivative of the function f (x) is usually denoted by f 00 (x) or,
using Leibniz’s notation:
 
d dfdx(x) d2 f
f 00 (x) = (x) = 2 (x)
dx dx

17 / 26
Example 5
In a number of macroeconomic applications the production function is written
as:
Y = f (K) = AK a where a ∈ (0, 1)
Then
f 0 (K) = aAK a−1 > 0 and f 00 (K) = a(a − 1)AK a−2 < 0
in other words, Y is increasing in K, but the slope of Y is decreasing in K.
Economists say that the production process is one of a diminishing marginal
product.

18 / 26
 f ′′ (x) ≤ 0 on I ⇐⇒ f ′ is decreasing on I
Curvature
The equivalence in (1) is illustrated in Fig. 1. The slope of the tange
as x increases. On the other hand, the slope of the tangent to the graph
Convexity as x increases. (Place a ruler as a tangent to the graph of the functi
along 0the curve from left to right, the tangent rotates counterclockwi
If f 00 (x) ≥ 0 on I ⇐⇒ f is
in Fig. 2.)
increasing on I ⇐⇒ Geometrically f is convex in
I.

y y

y ! f (x)

x
Figure 1 The slope of the tangent increases Figure 2 The slope of
as x increases. f ′ (x) is increasing as x increases. g ′ (x) is

We introduce the following definitions, assuming that f is continuo

19 / 26
 f ′′ (x) ≤ 0 on I ⇐⇒ f ′ is decreasing on I (2)

) is illustrated in Fig. 1. The slope of the tangent, f ′ (x), is increasing

other hand, the slope of the tangent to the graph in Fig. 2 is decreasing
a ruler as a tangent to the graph of the function. As the ruler slides
Concavity
eft to right,
00 the tangent rotates0 counterclockwise in Fig. 1, clockwise
If f (x) ≤ 0 on I ⇐⇒ f is decreasing on I ⇐⇒ geometrically f is concave
on I

y ! f (x)
y ! g(x)

x x
of the tangent increases Figure 2 The slope of the tangent decreases
is increasing as x increases. g ′ (x) is decreasing

wing definitions, assuming that f is continuous in the interval I and

20 / 26
Example 6
The function f (x) = x2 − 2x + 2 is convex

f 0 (x) = 2x − 2 ⇒ f 00 (x) > 0

21 / 26
Special functions

1. Exponential functions
2. Natural logarithm

22 / 26
Exponential functions

The exponential number e ≈ 2.71828 has the property that

f (x) = ex =⇒ f 0 (x) = ex > 0.

The above property implies that the n-th order derivative of the natural
exponential function is also equal to the function itself:

f (n) (x) = ex > 0

In particular, the natural exponential function must be increasing and convex.

23 / 26
Let a > 0. The function f (x) = ax is also an exponential function. In order to
take the derivative remember that the log natural is the inverse function of the
natural exponential function, hence

eln(a) = ln(ea ) = a

Thus, we have
ax = (eln(a) )x = eln(a)x
To get the derivative just apply the chain rule

(ax )0 = (eln(a)x ) ln(a) = ax ln(a)

Reminder: Applying the chain rule to eu(x) yields:

(eu(x) )0 = eu(x) u0 (x)

24 / 26
Logarithmic Functions
We use the equality x = eln(x) to find the derivative of the ln function.
Rule
h0 (x)
(ln h(x))0 =
h(x)

Proof
We know: h(x) = eln(h(x))
1. The derivative on the left hand side is just h0 (x)
2. The derivative on the right hand side is

(eln(h(x)) )0 = eln(h(x)) (ln(h(x))0 = h(x)(ln(h(x))0

so we can write:
h0 (x)
h0 (x) = h(x)(ln(h(x))0 ⇒ (ln h(x))0 =
h(x)
25 / 26
Logarithmic Differentiation

Consider the function f (x) = xx . Both the base and the exponent are functions,
so none of the rules discussed so far can be directly applied.

ln f (x) = x ln x

Then differentiating both sides yields

f 0 (x) x
= ln x +
f (x) x
0 x
f (x) = x (ln x + 1)

26 / 26
Derivatives in Use
Mathematics I
Lectures 3-4

Dr. Clemens Buchen

1 / 33
Plan of the lecture

I Implicit Differentiation
I Linear and quadratic approximations
I Differential
I Elasticities
I Limits, Continuity
Based on chapter 7

2 / 33
 Consider the following equation in x and y,
Implicit Differentiation
xy = 5 (∗)
Example 1
If x = 1, then y = 5. Also, x = 3 gives y = 5/3. And x = 5 gives y = 1. In general,
Consider the function:
for each number x ̸= 0, there is a unique number5 y such that the pair (x, y) satisfies the
y(x)y=implicitly as a function of x. The graph of
equation. We say that equation (∗) defines x
equation (∗) for x > 0 is shown in Fig. 1.
This function has a graph:
y
5
xy ! 5
4

1 2 3 4 5 x
Figure 1 xy = 5, x > 0

3 / 33
Suppose we interpret y as an implicit function of x. The definition then implies

x · y(x) = 5

The implicit equation x · y(x) = 5 must hold for all values of x. As a result, the
derivatives on both sides of the equation must also be the same.

y(x)
y(x) + x · y0 (x) = 0 ⇐⇒ y0 (x) = −
x
Substitution of y(x) yields y0 (x) = −5/x2 .
Alternatively, we could directly take the derivative of y = 5/x. :

5
y0 (x) = −
x2

4 / 33
Example 2
Consider the equation
y3 + 3x2 y = 13
What is the slope of the graph at the point (2, 1)?

5 / 33
I Interpreting y as a function of x yields:

[y(x)]3 + 3x2 y(x) = 13

I Differentiating and solving for y0 (x), we have:

3[y(x)]2 y0 (x) + 6xy(x) + 3x2 y0 (x) = 0

6xy(x)
⇐⇒ y0 (x) = −
3[y(x)]2 + 3x2

2xy(x)
⇐⇒ y0 (x) = −
[y(x)]2 + x2

Thus at the point (2, 1), we have

2·2·1 4
y0 (2) = − 2 2
=−
1 +2 5
6 / 33
The method of implicit differentiation

If two variables x and y are related by an equation, to find y0 :

1. Differentiate each side of the equation w.r.t. x, considering y as a
function of x. (Usually, you will need the chain rule).
2. Solve the resulting equation for y0 .

The Second Derivative of Implicit Functions

Repeat the procedure keeping in mind that you already know y0 . Try this with
the function in Example 1

7 / 33
An economic example

Example 3 (The multiplier effect)

In the standard macro model, national income is assumed to satisfy an
equation of the form:

Y = C + I where C = f (Y) with 0 < f 0 (Y) < 1

Work through one more example in the textbook, chapter 7.2

8 / 33
Linear Approximations
A linear approximation around the point a is given by the formula

f (x) ≈ f (a) + f 0 (a)(x − a)

9 / 33
The following will be useful below.
Example 4
In reference to the formula suppose we replace a by a generic x and we call
the change x − a = dx.
Then the linear approximation is given by:

f (x + dx) ≈ f (x) + f 0 (x)dx

10 / 33
The Differential of a Function

The term f 0 (x)dx is called the differential of y = f (x) and is denoted either by
dy or df :
dy = f 0 (x) dx
Suppose x changes by dx. Then the true change in y denoted ∆y is:

∆y = f (x + dx) − f (x)
≈ f (x) + f 0 (x)dx − f (x)
≈ f 0 (x)dx = dy

Obviously, the approximation is only valid for small changes in x.

11 / 33
Geometric representation of dy = f 0 (x) dx and ∆y

12 / 33
Example 5
Consider a simple macro model:

Y = C + I where C = f (Y) with 0 < f 0 (Y) < 1

The differentials are:

dY = dC + dI and dC = f 0 (Y)dY

Substituting:
1
dY = f 0 (Y)dY + dI ⇐⇒ dY = dI
1 − f 0 (Y)

13 / 33
Higher-Order approximations
We can extend the logic of the linear approximation. Intuitively, if we include
information about higher-order derivatives the approximation becomes better.
This results in Taylor’s formula. We skip it here, but you know where to find
it if you need it!
√
Example for f (x) = 3 x

14 / 33
Elasticities

Let D(P) denote the demand function. Then D0 (P) measures the reduction in
demand if price increase by e1.
This alone is problematic:
1. The concept ignores size effects: Raising the price of coffee by 1e is not
the same as raising the price of cars by 1e.
2. The concept ignores measurement scale effects. For example whether
quantities are measured in kg or pounds, or prices in e or $ would
significantly affect the result.

15 / 33
Again consider the demand function D(P). Then, the following term
measures the % change in demand if price varies by ∆P:

D(P + ∆P) − D(P) ∆D

=
D(P) D
The elasticity measures the % change of the dependent variable with respect
to a % variation of the independent variable. In the example of the demand
function, we obtain:

∆D
D Percentage change of the quantity
Price elasticity of demand = =
∆P
P
Percentage change of the price

16 / 33
We can rewrite the equality

∆D ∆P ∆D P D(P + ∆P) − D(P) P

/ = · = ·
D P ∆P D ∆P D
∆D dD
Finally, letting ∆P converge to zero implies ∆P → dP .

dD P
εPD (P) = ·
dP D

17 / 33
The General Definition of Elasticity

Elasticity of a differentiable function f at the point x with f (x) 6= 0:

x df x
εxf (x) = f 0 (x) = (x)
f (x) dx f (x)

Example 6
Suppose demand is given by D(p) = Ap−1
p
⇒ εpD = −Ap−2 · = −1
Ap−1
Interpretation: if the price increases by 1%, the quantity demanded decreases
by 1%.
This is the special case of an iso-elastic demand function.

18 / 33
Terminology
Consider the function f (x). Economists use the following terminology:
I If |εxf (x)| > 1, then f is elastic at x.
I If |εxf (x)| = 1, then f is unit elastic at x.
I If |εxf (x)| < 1, then f is inelastic at x.
I If |εxf (x)| = 0, then f is completely inelastic at x.

19 / 33
More on elasticities

Elasticities as logarithmic derivatives

dy x d(ln y)
εxy = =
dx y d(ln x)
Can you prove it?

The elasticity of the inverse is equal to the inverse of the elasticity

If y = f (x) has an inverse x = g(y), then g0 (y) = 1
f 0 (x) . Therefore:

g(y) y 1 f (x) 1
εy = g0 (y) = = f (x)
g(y) f 0 (x) x εx

20 / 33
Rules for Limits
Suppose the following holds:

lim f (x) = A and lim g(x) = B

x→a x→a

Then:
a.
lim (f (x) ± g(x)) = A ± B
x→a

b.
lim (f (x) · g(x)) = A · B
x→a
c.
f (x) A
lim = if B 6= 0
x→a g(x) B
d.
lim (f (x))r = Ar if Ar is defined
x→a

21 / 33
 write
1
One-sided
→limits
∞ as x → −2
(x + 2)2

y y
5
B
4
3 y " f (x)

2 A

!2 !1 1 x a x

→ ∞ as x → −2 Figure 2 lim f (x) does not exist

x→a

lim f (x) = A
a number, so ∞ is not a limit.) x→a−

lim f (x) = B
x→a+

mits
22 / 33
This leads to a necessary and sufficient condition for the existence of a limit:
 
lim f (x) = A ⇐⇒ lim− f (x) = A and lim+ f (x) = A
x→a x→a x→a

23 / 33
indicates that f (x) can be made arbitrarily close to B by making x a sufficiently large
negative number. The two limits are illustrated in Fig. 4. The horizontal line y = A is a
Limits atasymptote
(horizontal) infinityfor the graph of f as x tends to ∞, whereas y = B is a (horizontal)
asymptote for the graph as x tends to −∞.

y
A

Figure 4 y = A and y = B are horizontal asymptotes

lim f (x) = A
as x → ∞ and as x → −∞:
Examine the following functions x→∞
lim f (x) = B
3x 2 + x − 1 x→−∞
1 − x5
(a) f (x) = (b) g(x) =
x2 + 1 x4+x+1

Solution:
24 / 33
Example 7
Examine the following function as x → ∞ and x → −∞:

3x2 + x − 1
f (x) =
x2 + 1
Divide by the largest power of x:

3 + 1x − x12
f (x) =
1 + x12

We conclude that f (x) → 3 as x → ∞ and x → −∞.

25 / 33
Note: We have to be careful when we combine functions:


 f (x) + g(x) → ∞
f (x)g(x) → ∞

f (x) → ∞ and g(x) → ∞ as x → a =⇒

 f (x) − g(x) →?
f (x)/g(x) →?


Example 8
Let f (x) = x−2 and g(x) = x−4 . As x → 0 , both tend to infinity. What about
f − g, g − f , f /g, and g/f ?
x2 −1
f (x) − g(x) = x4

1−x2
g(x) − f (x) = x4

f (x)/g(x) = x2

g(x)/f (x) = x−2

26 / 33
Continuity
I A function is continuous if small changes in the independent variable
induce small changes in the function value.
I Geometrically, a function is continuous on an interval if its graph is
connected, i.e. it has no breaks.

27 / 33
Continuity in Terms of Limits

Analytically, a function is said to be continuous at the point x = a if

f (a) = limx→a f (x)
Hence, to verify that f is continuous at x = a, the following three conditions
should hold:
1. The function f must be defined at x = a
2. The limit of f (x) as x tends to a must exist
3. This limit must be exactly equal to f (a)

28 / 33
Theorem
Suppose f and g are continuous at a, then
(a) f + g and f − g are continuous at a
(b) f · g and f /g (if g(a) 6= 0) are continuous at a
(c) [f (x)]r is continuous at a if [f (a)]r is defined (where r any real number)
(d) If f is continuous and has an inverse on the interval I, then its inverse f −1
is continuous on f (I).
As a result, functions which are constructed from continuous functions by
addition, subtraction, multiplication, division (except by zero) and
composition are continuous at all points where it is defined.

29 / 33
Continuity and Differentiability

30 / 33
Example 9
Let f (x) = |x|. Clearly it is continuous at x = 0. However
Taking the limit from the left, we get the left derivative:

f (a + h) − f (a)
lim− = −1
h→0 h

Taking the limit from the right yields:

f (a + h) − f (a)
lim+ = +1
h→0 h

In other words, the limit limh→0 f (a+h)−f

h
(a)
does not exist.

31 / 33
L’Hopital’s Rule

Sometimes we look for the limit a quotient at a point in which both numerator
and denominator tend to 0. Specifically, suppose we need

f (x)
lim
x→a g(x)

but we have limx→a f (x) = 0 and limx→a g(x) = 0. What next?

Theorem (l’Hopital rule):

Assuming limx→a f (x) = 0 and limx→a g(x) = 0 and both functions are
differentiable at a, then:
f (x) f 0 (a)
lim = lim 0
x→a g(x) x→a g (a)

32 / 33
Justification: Using a linear approximation we can write

f (x) f (a) + f 0 (a)(x − a) f 0 (a)

≈ =
g(x) g(a) + g0 (a)(x − a) g0 (a)

Example 10
x
Suppose f (x) = e x−1 . The function is obviously not well defined at the point
x = 0. What is the limit of f at zero? Directly applying l’Hôpital rule yields:
ex − 1 ex
lim = lim =1
x→0 x x→0 1

33 / 33
 Optimization
Mathematics I
Lectures 5-6

Dr. Clemens Buchen

1 / 29
Plan of the lecture

I Idea of extreme points

I First-derivative test
I Second-derivative test
I Global versus local extreme points
I Extreme value theorem
Based on chapter 8

2 / 29
Extreme points

Definition
Points in the domain of a function where it reaches its largest and its smallest
values are referred to as maximum and minimum points, or extreme points.
Thus, if f (x) has domain D, then

c ∈ D is a maximum point for f ⇔ f (x) ≤ f (c) for all x ∈ D

d ∈ D is a minimum point for f ⇔ f (x) ≥ f (d) for all x ∈ D

3 / 29
First-Order Condition

Theorem
Suppose that a function f is differentiable in an interval I and that c is an
interior point of I. For x = c to be a maximum or minimum point for f in I, a
necessary condition is that it satisfies the equation

f 0 (c) = 0 first-order condition or f.o.c.

We say that c is a critical or stationary point of the function.

4 / 29
Proof
Suppose x =c is a maximum (similar logic for a minimum), but that contrary
to the claim f 0 (c) > 0 (similar for <)
I Since c ∈ int(D) it is possible to marginally increase c (reduce c).
I For sufficiently small h, it must be that f (c + h) > f (c), which leads to a
contradiction (or for a min f (c + h) < f (c)).

5 / 29
Note 1
f 0 (x) = 0 is a necessary condition for a differentiable function to have a
maximum or minimum at an interior point x in its domain. The condition is
not sufficient.

6 / 29
Note 2
f 0 (x) = 0 is not sufficient. It does not tell you whether you are at a (local)
maximum, at a (local) minimum or at an inflection point.

7 / 29
First-derivative test

Suppose f (x) is differentiable in an interval I and that it has a single stationary

point, x = c., i.e. f 0 (c) = 0.
1. Suppose f 0 (x) ≥ 0 for all x in I such that x ≤ c, whereas f 0 (x) ≤ 0 for all
x in I such that x ≥ c. Then f (x) is increasing to the left of c and
decreasing to the right of c. It follows that f (x) ≤ f (c) for all x ≤ c, and
f (c) ≥ f (x) for all x ≥ c.
Hence, x = c is a maximum point for f in I.

8 / 29
9 / 29
2. Alternatively, suppose f 0 (d) = 0, with
I f 0 (d) ≤ 0 for all x in I such that x ≤ d
I f 0 (d) ≥ 0 for all x in I such that x ≥ d.
Then x = d is a minimum point for f in I.

10 / 29
11 / 29
Example 1
Suppose the profit of a firm as a function of the quantity produced is given by
the function
π(q) = 2q − q2

12 / 29
Second-derivative test

I Suppose that f is concave with f 00 (x) ≤ 0 for all x in an interval I. Then

f 0 (x) is decreasing in I.
I If f 0 (c) = 0 at an interior point c of I, then it must be that f 0 (x) ≥ 0 to the
left of c, while f 0 (x) ≤ 0 to the right of c. This implies that the function
itself is increasing to the left of c and decreasing to the right of c.
I Thus, the point x = c must be a maximum point for f in I. We obviously
get a corresponding result for a minimum of a convex function.

13 / 29
Theorem: Maximum/Minimum for Concave/Convex Functions
Suppose f is a concave (convex) function in an interval I. If c is a stationary
point for f in the interior of I, then c is a maximum (minimum) point for f in I.

14 / 29
Example 2
The function f (x) = ex−1 − x takes a minimum at the point x = 1.
1. The function is convex: f 00 (x) = ex−1 > 0
2. The function has a stationary point at x = 1.

f 0 (x) = ex−1 − 1 = 0 ⇒ x=1

15 / 29
Example 3
A monopolist is faced with the inverse demand function P(Q) (i.e. P(Q)
denotes the price when output is Q). Suppose the monopolist’s cost function
is C(Q) = kQ.
I The monopolist’s profit is

π(Q) = P(Q)Q − kQ

I The necessary condition for profit maximization yields:

π 0 (Q∗ ) = P0 (Q∗ )Q∗ + P(Q∗ ) − k = 0

⇒ P0 (Q∗ )Q∗ + P(Q∗ ) = k

I Note that implicitly Q∗ is a function of k! How does the optimal quantity

change if the per unit costs increase? What is your intuition? Verify your
intuition using implicit differentiation.
See section 8.3 for more economic examples.
16 / 29
Extreme Value Theorem

If f is a continuous function over a closed bounded interval [a, b], then there
exists
I a point d ∈ [a, b] where f has a minimum,
I a point c ∈ [a, b] where f has a maximum,
I Altogether, there are two points c, d ∈ [a, b], such that

∀x ∈ [a, b], f (d) ≤ f (x) ≤ f (c)

Observe that the conditions are sufficient, but not necessary conditions for an
extreme point.

17 / 29
How to Search for Maxima/Minima
Every extreme point must belong to one of the following three different sets:
1. Interior points in I where f 0 (x) = 0.
2. End points of I
3. Interior points in I where f 0 does not exist.

18 / 29
Example 4
Find the maximum and minimum values for

f (x) = 3x2 − 6x + 5 with x ∈ [0, 3]

1. Differentiating yields

f 0 (x) = 6x − 6 = 0 ⇒ x = 1 and f (1) = 2

2. At the end points, we have

f (0) = 5, f (3) = 27 − 18 + 5 = 14

3. The function is everywhere differentiable.

Thus, the minimum is reached at the point x = 1 and a maximum at the point
x = 3.
See Example 8.5.2 for an economic application of the extreme value theorem.

19 / 29
Global extreme points

Suppose a function is concave and has a stationary point. Then the stationary
point is a global maximum.

Suppose a function is convex and has a stationary point. Then the stationary
point is a global minimum.

20 / 29
Global and local extreme points

21 / 29
The Second-Derivative Test for Local Extreme Points

Example 5
1 1 2
f (x) = x3 − x2 − x + 1
9 6 3
Check the first-order condition
1 2 1 2
f 0 (x) = x − x−
3 3 3
1 2 1
= (x − x − 2) = (x + 1)(x − 2)
3 3
It follows that there two stationary points:

x1 = −1
x2 = 2

22 / 29
The second derivative is
2 1
f 00 (x) = x −
3 3
Evaluate the second derivative at the stationary points:

f 00 (−1) = −1
f 00 (2) = 1

We find that the function is locally concave around x = −1 and locally convex
around x = 2.
Note that the first-derivative test could be applied here as well. Try it.

23 / 29
Second-Derivative Test

Let f be a twice differentiable function in an interval I , and let c be an interior

point of I. Then:
a. f 0 (c) = 0 and f 00 (c) < 0 ⇒ x = c is a strict local maximum point.
b. f 0 (c) = 0 and f 00 (c) > 0 ⇒ x = c is a strict local minimum point.
Intuition: Consider case (a). For small positive h we have
I f 0 (c − h) > 0
I f 0 (c + h) < 0
I At the point c we must have a local maximum.

24 / 29
Example 6
Consider the function f (x) = x2 ex

1. The first derivative is:

f 0 (x) = 2xex + x2 ex
= (2 + x )xex

which is 0 at x = −2 or x = 0.
2. The second derivative is:

f 00 (x) = xex + (2 + x)ex + (2 + x)xex

= (2 + 4x + x2 )ex

25 / 29
Thus, at the point
x = −2 the second derivative is

f 00 (−2) = 2 + 4(−2) + (−2)2 e−2 = −2e−2 < 0

 

and the function has a local maximum, and at

x = 0 the second derivative is

f 00 (0) = 2 + 4 · 0 + 02 e−0 = 2 > 0

 

and the function has a local minimum.

26 / 29
Inflection Points

Example 7
Consider the function
1 1 1
f (x) = x2 − x3 − x
2 6 2
We have
1 1
f 0 (x) = x − x2 −
2 2
00
f (x) = 1 − x

It is straightforward to show that f 0 (1) = 0, but then also f 00 (1) = 0.

27 / 29
Definition
The point c is called an inflection point for the function f if there exists an
interval (a, b) about c such that:
I f 00 (x) ≥ 0 in (a, c) and f 00 (x) ≤ 0 in (c, b), or
I f 00 (x) ≤ 0 in (a, c) and f 00 (x) ≥ 0 in (c, b).

Let f be a function with a continuous second derivative in an interval I, and let

c be an interior point of I.
1. If c is an inflection point for f , then f 00 (c) = 0.
2. If f 00 (c) = 0 and f 00 changes sign at c, then c is an inflection point for f .

28 / 29
 1. Because f 00 (x) ≤ 0 on one side of c and f 00 (x) ≥ 0 on the other, and
because f 00 is continuous, it must be true that f 00 (c) = 0.
2. If f 00 changes sign at c, then c is an inflection point for f , according to the
definition of an inflection point.
Note: f 00 (c) = 0 is a necessary condition for c to be an inflection point. It is
not a sufficient condition, however, because f 00 (c) = 0 does not imply that f 00
changes sign at x = c.

29 / 29
 Integration
Mathematics I
Lectures 7-8

Dr. Clemens Buchen

1 / 37
Plan of the lecture

I Indefinite Integrals
I Definite Integrals (Areas)
I Differentiation w.r.t. limits
I Some methods of integration
Based on chapter 9

2 / 37
Indefinite Integrals

I Suppose we do not know a function F : R 7→ R, but that F 0 (x) = x2 .

What can we say about F?

I Consider the function:

1
G(x) = F(x) − x3
3
Observe that G0 (x) = 0 for all x =⇒ G(x) = C for all x, where C is an
arbitrary constant.
I Thus
1
F 0 (x) = x2 ⇐⇒ F(x) = x3 + C
3

3 / 37
Indefinite Integral
Definition

For F 0 (x) = f (x)

Z
f (x) dx = F(x) + C (1)

I Symbol : integral sign

R

I Function f (x) in (1): integrand

I dx : x is the variable of integration
I C : constant of integration.

4 / 37
Integration and differentiation cancel each other out:
d
Z
f (x) dx = f (x)
dx
Z
F 0 (x) dx = F(x) + C

5 / 37
Some Important Integrals

I Important integration formulas which follow immediately from the rules

for differentiation.
I The indefinite integral of any power of x (except x−1 ) is obtained by
increasing the exponent of x by 1, then dividing by the new exponent,
and finally adding a constant of integration
1 a+1
Z
xa dx = x +C (a 6= −1) (2)
a+1

6 / 37
Example 1
1 1+1 1
Z Z
x dx = x1 dx = x + C = x2 + C
1+1 2

Example 2
1 1 1
Z Z
dx = x−3 dx = x−3+1 + C = − 2 + C
x3 −3 + 1 2x

Example 3
Z
√ Z
1 1 1 +1 2 3
x dx = x 2 dx = 1
x2 +C = x2 +C
2 +1 3

7 / 37
In the case a = −1 formula (2) is not valid.
1. Remember:
1
(ln x)0 = for all x > 0
x
2. Alternatively, suppose x < 0 then ln(−x) is defined and

1 1
(ln(−x))0 = (−1) =
−x x
3. Moreover, recall that |x| = x when x ≥ 0 and |x| = −x when x < 0. Thus,
we get:
1
Z
dx = ln |x| + C
x

8 / 37
Exponential function

1
Z
eax dx = eax + C (a 6= 0) (3)
a
For a > 0 we can write
h ix
ax = e(ln a) = e(ln a)x

Thus as a special case of (3), when ln a 6= 0 because a 6= 1, we obtain

1 x
Z
ax dx = a +C (a > 0 and a 6= 1)
ln a

9 / 37
Some General Rules
The differentiation rules

(aF(x))0 = aF 0 (x) and (F(x) + G(x))0 = F 0 (x) + G0 (x)

imply the following integration rules

Z Z
af (x) dx = a f (x) dx (a is a constant)
Z Z Z
[f (x) + g(x)] dx = f (x) dx + g(x) dx

More generally, we have:

Z Z Z
[a1 f1 (x) + · · · + an fn (x)] dx = a1 f1 (x) dx + · · · + an fn (x) dx

10 / 37
Area and Definite Integrals

Suppose f : R 7→ R is a continuous function. Consider the area under f in the

interval [a, b].
11 / 37
Let t be an arbitrary point in [a, b], and let A(t) denote the area under the curve
y = f (x) over the interval [a, t].

12 / 37
For all ∆t > 0 : f (t) ∆t ≤ A(t + ∆t) − A(t) ≤ f (t + ∆t) ∆t
A(t + ∆t) − A(t)
⇐⇒ f (t) ≤ ≤ f (t + ∆t)
∆t

13 / 37
As ∆t → 0
The interval [t, t + ∆t] shrinks to the single point t, and by continuity of f

f (t + ∆t) → f (t)

Thus, we have
A(t + ∆t) − A(t)
lim = f (t)
∆t→0 ∆t
Therefore A(t), which measures the area under the graph of f over the interval
[a, t], is differentiable, with

A0 (t) = f (t) for all t in (a, b)

14 / 37
Suppose that F(x) is an arbitrary indefinite integral of f (x).
Then A(x) = F(x) + C for some constant C.
Recall that A(a) = 0. Hence, 0 = A(a) = F(a) + C, so C = −F(a). Therefore,
Z
A(x) = F(x) − F(a) where F(x) = f (x) dx

Example 4
Calculate the area under the parabola f (x) = x2 over the interval [0, 1].

15 / 37
Definite Integral

Definition
Z b b

f (x) dx =  F(x) = F(b) − F(a)
a a
where F is any function satisfying F 0 (x) = f (x) for all x ∈ (a, b).
Numbers a and b : lower and upper limit of integration.

Although the notation for definite and indefinite

Rb
integrals is similar, the two
integrals Rare entirely different. In fact, a f (x) dx denotes a single number,
whereas f (x) dx represents any one of the infinite set of functions all having
f (x) as their derivative.

16 / 37
Properties of Definite Integrals
If f , g are continuous functions in an interval that contains a, b, and c, and
α, β are real numbers, then
Z b Z a
f (x) dx = − f (x) dx
a b
Z a
f (x) dx = 0
a
Z b Z b
αf (x) dx = α f (x) dx
a a
Z b Z c Z b
f (x) dx = f (x) dx + f (x) dx
a a c
Z b Z b Z b
(αf (x) + β g(x)) dx = α f (x) dx + β g(x) dx
a a a

17 / 37
Differentiation w.r.t. the Limits of Integration

Example 5
Consider the definite integral: Z t
x2 dx
0
Suppose we wanted to find the derivative w.r.t. t. We know that
Z t
1
x2 dx = t3
0 3
Therefore: Z t  
d 2 d 1 3
x dx = t = t2
dt 0 dt 3

18 / 37
More generally, suppose a(t) and b(t) are differentiable and f (x) is
continuous, then Z b(t)
f (x) dx = F(b(t)) − F(a(t))
a(t)

Then:
Z b(t)
d
f (x) dx = F 0 (b(t))b0 (t) − F 0 (a(t))a0 (t)
dt a(t)

= f (b(t))b0 (t) − f (a(t))a0 (t)

In our example a(t) = 0 and b(t) = t.

19 / 37
Note: If t is additionally also in the integrand we need an additional term:
Leibniz’s Formula
Suppose
Z b(t)
G(t) = f (x, t) dx
a(t)

Then Z b(t)
∂ f (x, t)
G0 (t) = F 0 (b(t))b0 (t) − F 0 (a(t))a0 (t) + dx
a(t) ∂t

Look it up here if you need it (not needed for the exam in the first year...)

20 / 37
Example 6
Assume that at time t = 0 we start extracting oil from a well that contains K
barrels of oil.

x(t) = number of barrels of oil that is left at time t,with x(0) = K

x(t) is a decreasing function of t

The amount of oil that is extracted in a time interval [t, t + ∆t] , where ∆t > 0,
is x(t) − x(t + ∆t)
Extraction per unit of time is

x(t) − x(t + ∆t) x(t + ∆t) − x(t)

=− (∗)
∆t ∆t
Assuming that x(t) is differentiable, the limit as ∆t approaches zero in the
fraction (∗) is equal to −x0 (t). Often the derivative w.r.t. time is written as
ẋ(t).

21 / 37
Letting u(t) denote the rate of extraction at time t

ẋ(t) = −u(t) with x(0) = K

Actually, this is an example of a differential equation. This one has the

solution Z t
x(t) = − u(τ) dτ
0

Note: we will not go into the details of differential equations. If you are interested,
you can read chapter 9.8

22 / 37
Integration by Parts

Consider the following integral:

Z
xex dx

In general, because the derivative of a product is not the product of the

derivatives, the integral of a product is not the product of the integrals.

23 / 37
Recall the product rule:

(f (x)g(x))0 = f 0 (x)g(x) + f (x)g0 (x)

Taking the indefinite integral of each side and using the rule for integrating a
sum gives: Z Z
f (x)g(x) = f 0 (x)g(x) dx + f (x)g0 (x) dx

where the constants of integration are implicit in the indefinite integrals on the
right-hand side of this equation. Rearranging:
Integration by Parts:
Z Z
f (x)g0 (x) dx = f (x)g(x) − f 0 (x)g(x) dx

24 / 37
Example 7
Use integration by parts to evaluate xex dx.
R

Z Z
x x
x · e dx =
|{z} ·e −
x|{z} 1 · ex dx
|{z}
f (x)·g0 (x) f (x)·g(x) f 0 (x)·g(x)
Z
= xex − ex dx
= xex − ex + C

Example 8
Evaluate Z
ln x dx

25 / 37
Integration by Substitution

Example 9
Z
50
x2 + 10 2x dx

Define u = x2 + 10 and du = 2x dx and write down and solve the integral

1 51
Z
u50 du = u +C
51
Hence,
1 2
Z
50
51
x2 + 10 2x dx = x + 10 + C
51
 0
1 51
50
x2 + 10 + C = x2 + 10 2x, so we have indeed


Let us check: 51
confirmed the result.

26 / 37
Integration by Substitution
Z Z
f (g(x))g0 (x)dx = f (u)du with u = g(x)

R
Method for finding a complicated integral G(x)dx
1. Choose a part of G(x) and introduce it as the new variable u = g(x).
2. Compute the differential du = g0 (x)dx.
R R
3. Substitute u and du into G(x)dx to get f (u)du
R
4. Evaluate f (u)du = F(u) + C
5. Resubstitue g(x)

27 / 37
In case of a definite interval it is sometimes quicker to reformulate the upper
and lower limits of integration:
Z b Z g(b)
0
f (g(x))g (x)dx = f (u)du with u = g(x)
a g(a)

28 / 37
Example 10
Z a
2
xe−cx dx
0

1
1. Substitute u = −cx2 , which gives du = −2cx dx, or x dx = − 2c du.
2. Hence, the new integral is given by
Z a
1 u
− e du
0 2c
.
3. Next we adjust the limits: u(0) = 0 and u(a) = −ca2 , so:
 2
1 −ca u
Z −ca2
1 1 2
− u
e du = −  e = − (e−ca − 1)
2c 0 2c 0 2c

29 / 37
Infinite Intervals of Integration

Consider Example 10 again. We have established that

Z a
2 1 2
xe−cx dx = (1 − e−ca )
0 2c
As we let a tend to infinity (and c is a positive number) the last term tends to
zero or
1
Z ∞
2
xe−cx dx =
0 2c

30 / 37
Example 11
Consider the density function of the exponential distribution:

f (x) = λ e−λ x with x ≥ 0; λ > 0

Show that the area below the graph over [0, ∞) is equal to 1.
 Z b b  
−λ x
dx =  −e−λ x = −e−λ b + 1

λe
0 0

As b → ∞, e−λ b approaches 0. Therefore,

Z ∞ Z b
−λ x
λe dx = lim λ e−λ x dx = lim (−e−λ b + 1) = 1
0 b→∞ 0 b→∞

32 / 37
This allows us to formulate a general rule.

Suppose f is a function that is continuous for all x ≥ a. If the limit of this

integral as b ≥ a tends to infinity exists, f is integrable over [a, ∞):
Z ∞ Z b
f (x)dx = lim f (x)dx
a b→∞ a

I If the limit exists, we say that a∞ f (x)dx converges.

R

I If the limit does not exist, it is said to diverge.

33 / 37
Integrals of Unbounded Functions
Example 12

Consider the function f (x) = √1 . We are interested in the integral

x

2
Z 2
1  √ √ √
√ dx =  2 x = 2 2 − 2 h
h x h
Now let h → 0+ . The integral becomes
Z 2
1 √
√ dx = 2 2
0 x

The improper integral of the area below f converges.

More generally, suppose that f is a continuous function in the interval (a, b],
but f is not defined at x = a. Then we can define
Z b Z b
f (x)dx = lim+ f (x)dx
a h→0 a+h

If the limit exists, the improper integral converges.

35 / 37
Example 13
R +1 1
Consider the integral −1 x2 dx. A naive (and wrong!) evaluation could lead us
to calculate 1
Z +1
−2
 1
x dx =  − = −2
−1 −1 x
1
This, however, cannot be true because x2
is never negative. The problem
arises because limx→0 x12 = +∞.

36 / 37
 Z +1 Z 0 Z +1 Z 0−h Z 1
−2 −2 −2 −2
x dx = x dx + x dx = lim x dx + lim x−2 dx
−1 −1 0 h→0 −1 k→0 0+k

Finding the definite integral:

0−h 1    
 1  1 1 1
lim  − + lim 
 − = lim − − 1 + lim −1 + =∞
h→0 −1 x k→0 0+k x h→0 0−h k→0 0+k

37 / 37
Functions of Many Variables
Mathematics I
Lectures 9-10

Dr. Clemens Buchen

1 / 34
Plan of the lecture

I Definition of a function with two variables

I Domain
I Partial derivatives
I Geometric interpretation
I n dimensions
I Partial elasticities
Based on chapter 11

2 / 34
A function of two variables

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

I Let z denote the value of f at (x, y), so z = f (x, y).

I We call x and y the independent variables, or the arguments of f ,
I z is called the dependent variable.
I The domain of the function f is then the set of all possible pairs of the
independent variables.
I The range is the set of corresponding values of the dependent variable.

3 / 34
Example 1
Consider the function f :R × R 7−→ R with

f (x, y) = 2x + x2 y3 .

Example 2
A study of the demand for milk by R. Frisch and T. Haavelmo found the
relationship
m2.08
x = A 1.5 (A is a positive constant)
p

Example 3
The Cobb–Douglas function

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

Usually, one assumes that F is defined only for x > 0 and y > 0.
4 / 34
The Domain

I For functions of two variables x and y, the domain D is a set of points in

the xy-plane.
I Sometimes it is helpful to draw a graph of the domain in the xy-plane.

Example 4
√ √
What is the domain of the function F(x, y) = x−1+ y ?

5 / 34
Partial Derivatives with Two Variables

I Consider the function z = f (x, y) defined as

z = x3 + 2y2 . (1)

I Suppose first that y is held constant. Then 2y2 is constant. Really there is
only one variable now. Of course, the rate of change of z w.r.t. x is given
by
dz
= 3x2 .
dx
I Similarly, we can keep x fixed in (1) and examine how z varies as y
varies:
dz
= 4y.
dy

6 / 34
Notation
We write ∂ z/∂ x instead of dz/dx for the derivative of z w.r.t. x when y is held
fixed:
∂z ∂z
z = x3 + 2y2 =⇒ = 3x2 and = 4y
∂x ∂y
Other common notation to indicate the partial derivatives of z = f (x, y):

∂f ∂z ∂ f (x, y)
= = zx = fx = fx0 (x, y) = f10 (x, y) =
∂x ∂x ∂x
∂f ∂z ∂ f (x, y)
= = zy = fy = fy0 (x, y) = f20 (x, y) =
∂y ∂y ∂y

7 / 34
Formal Definition
∂ f (x, y) f (x + h, y) − f (x, y)
= lim
∂x h→0 h
∂ f (x, y) f (x, y + h) − f (x, y)
= lim
∂y h→0 h
All the rules for differentiation from the first lecture still apply here!

8 / 34
Example 5
Consider the function f (x, y) = x3 y + x2 y2 + x + y2 . What are its partial
derivatives?

9 / 34
Higher-Order Partial Derivatives

I If z = f (x, y), then ∂ f /∂ x and ∂ f /∂ y are called first-order partial

derivatives. They are, in general, again functions of two variables.
I From ∂ f /∂ x, provided this derivative is itself differentiable, we can
generate two new functions by taking the partial derivatives w.r.t. x and y.
I The four functions obtained by differentiating twice in this way are
called second-order partial derivatives of f (x, y). They are expressed as

∂ 2f ∂ 2f
   
∂ ∂f ∂ ∂f
= , = ,
∂x ∂x ∂ x2 ∂y ∂x ∂ y∂ x
∂ 2f ∂ 2f
   
∂ ∂f ∂ ∂f
= , = 2.
∂x ∂y ∂ x∂ y ∂y ∂y ∂y

10 / 34
Example 6
For the function f (x, y) = x3 y + x2 y2 + x + y2 , we obtain

∂f ∂f
= 3x2 y + 2xy2 + 1, = x3 + 2x2 y + 2y
∂x ∂y

11 / 34
Other notations for second-order partial derivatives:

∂ 2f 00 (x, y) or f 00 (x, y) or f (x, y) or f (x, y)

∂ x2
can also be written as f11 xx 11 xx

∂ 2f 00 (x, y) or f 00 (x, y) or f (x, y) or f (x, y)

∂ y∂ x can also be written as f12 xy 12 xy

12 / 34
Geometric Representation
The graph of z = f (x, y) defined over a domain D in the xy-plane.

13 / 34
Level curves

14 / 34
The graph of z = f (x, y) and one of its level curves c = f (x, y).
15 / 34
Example 7
Consider the function of two variables defined by the equation

z = x2 + y2 . (2)

Each level curve has the equation

x2 + y2 = c (3)

for some c ≥ 0. We see that these are circles in the xy-plane centred at the
√
origin and with radius c

16 / 34
Level curves for x2 + y2 = c
17 / 34
z = x 2 + y2
18 / 34
Geometric Interpretations of Partial Derivatives

19 / 34
Example 8

20 / 34
On the basis of this figure:
1. What are the signs of fx (x, y) and fy (x, y) at the points P and Q?
2. Use the figure to estimate fx (3, 1).
3. What is the largest value that f (x, y) can attain when x = 2, and for which
y value does this maximum occur?

21 / 34
Gradients
A vector containing the partial derivatives at a point (a, b) is called the
gradient or gradient vector:
f10 (a, b), f20 (a, b)

22 / 34
Functions of n variables

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

I n-vectors are often denoted by bold letters x = (x1 , x2 , . . . , xn ).

Definition
A function f of n variables x1 , . . . , xn with domain D is a rule that
assigns a specified number f (x) = f (x1 , . . . , xn ) to each n-vector x =
(x1 , . . . , xn ) in D.

23 / 34
Typical functions

Linear function:
f (x1 , x2 , . . . , xn ) = a1 x1 + a2 x2 + · · · + an xn + b
where a1 , a2 , . . . , an , and b are constants.

The general Cobb–Douglas function:

F(x1 , x2 , . . . , xn ) = Ax1a1 x2a2 · · · xnan (A, a1 , ..., an are constants; A > 0)

defined for x1 > 0, x2 > 0, . . . , xn > 0.

Log-linear function:
ln F = ln A + a1 ln x1 + a2 ln x2 + · · · + an ln xn

24 / 34
Euclidean n-Dimensional Space

I We call the set of all possible n-vectors (x1 , x2 , . . . , xn ) of real numbers

the Euclidean n-dimensional space, or n-space, and to denote it by Rn .
I If z = f (x1 , x2 , . . . , xn ) = f (x) represents a function of n variables, we
define the graph of f as the set of all points (x, f (x)) ∈ 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 .
I For z = z0 (a constant), the set of points in Rn satisfying f (x) = z0 is
called a level surface of f .

25 / 34
Partial Derivatives with More Variables

I If z = f (x) = f (x1 , x2 , . . . , xn ), then ∂ f /∂ xi , for i = 1, 2, . . . , n, means the

partial derivative of f (x1 , x2 , . . . , xn ) w.r.t. xi when all the other variables
xj (j 6= i) are held constant.
I So provided they all exist, there are n partial derivatives of first order,
one for each variable xi , i = 1, . . . , n. Other notation used for the
first-order partials of z = f (x1 , x2 , . . . , xn ) includes

∂f ∂z
= = ∂ z/∂ xi = z0i = fi0 (x1 , x2 , . . . , xn ) = fxi (x1 , x2 , . . . , xn ).
∂ xi ∂ xi

26 / 34
Example 9a
Consider the function f (x1 , x2 , x3 ) = 5x12 + x1 x23 − x22 x32 + x33 .
The partial derivatives are

f1 = 10x1 + x23 , f2 = 3x1 x22 − 2x2 x32 , f3 = −2x22 x3 + 3x32 .

27 / 34
For each of the n first-order partials of f , we have n second-order partials:

∂ 2f
 
∂ ∂f
= = fij00 = z00ij .
∂ xj ∂ xi ∂ xj ∂ xi

Here both i and j may take any value 1, 2, . . . , n, so altogether there are n2
second-order partials.
It is usual to display these second-order partials in an n × n matrix called the
Hessian matrix:
 00 00 (x) ... f 00 (x)

f11 (x) f12 1n
 f 00 (x) f 00 (x) ... f 00 (x) 
f 00 (x) = 
 :
21 22 2n .
: : 
00 (x) f 00 (x) ... f 00 (x)
fn1 n2 nn

28 / 34
Example 9b
Find the Hessian matrix of f (x1 , x2 , x3 ) = 5x12 + x1 x23 − x22 x32 + x33 .
00 00 00 3x22
   
f11 f12 f13 10 0
00
H =  f21 00
f22 00  =  3x2 6x x − 2x2
f23 −4x2 x3 
2 1 2 3
00 00 00
f31 f32 f33 0 −4x2 x3 −2x22 + 6x3

29 / 34
Young’s Theorem

Suppose that all the m−th-order partial derivatives of the function

f (x1 , x2 , . . . , xn ) are continuous. If any two of them involve differentiating
w.r.t. each of the variables the same number of times, then they are necessarily
equal.
In particular, for the case when m = 2,

∂ 2f ∂ 2f
= (i = 1, 2, . . . , n; j = 1, 2, . . . , n)
∂ xj ∂ xi ∂ xi ∂ xj

if both these partials are continuous.

30 / 34
Economic Applications

Example 10
Consider an agricultural production function

F(K, L, T) = AK a Lb T c with A > 0 and a, b, c ∈ (0, 1).

31 / 34
Partial Elasticities

If z = f (x, y), we define the (partial) elasticity of z w.r.t. x and y by

x ∂z y ∂z
εxz (x, y) = (x, y), εyz (x, y) = (x, y).
z ∂x z ∂y

I Where confusion is not possible, we often write εxz instead of εxz (x, y),
and εyz instead of εyz (x, y).
I The number εxz is (approximately) the percentage change in z caused by a
1% increase in x when y is held constant, and εyz has a corresponding
interpretation.

32 / 34
Example 11
Consider the Cobb Douglas function z = Axa yb .
x
εxz = Aaxa−1 yb
z
1
= Aaxa yb = a
(Axa yb )
y
εyz = Abxa yb−1
z
1
= Abxa yb = b
(Axa yb )

33 / 34
Partial Elasticities

General definition
If z = f (x1 , x2 , . . . , xn ) = f (x), we define the (partial) elasticity of z (or
of f ) w.r.t. xi as the elasticity of z w.r.t. xi when all the other variables
are held constant. Thus,
xi ∂ f (x) xi ∂ z ∂ ln z
εiz (x) = = = .
f (x) ∂ xi z ∂ xi ∂ ln xi

The intuition remains the same; εiz (x) is approximately equal to the
percentage change in z caused by a 1% increase in xi , at the point x, keeping
all the other xj (j 6= i) constant.

34 / 34
Tools for Comparative Statics
Mathematics I
Lectures 11-12

Dr. Clemens Buchen

1 / 36
Plan of the lecture

I Generalize some comparative statics tools for 2 and more variables

I Chain rule
I Linear approximations
I Differentials
I Implicit differentiation
I Homogeneity of functions
I Systems of equations
Based on chapter 12

2 / 36
A Simple Chain Rule

Example 1
Many models of economic growth regard output as a function of capital and
labour, both of which are themselves functions of time.
Suppose z is a function of x and y, with

z = F(x, y)

where x and y both are functions of a variable t, with

x = f (t) and y = g(t).

=⇒ z(t) = F(f (t), g(t)).

3 / 36
Will a small increase in t lead to an increase or a decrease in z?
dz dx dy
= Fx (x, y) + Fy (x, y)
dt dt dt

dz
The total derivative dt is the sum of the two contributions:

dx
Fx (x, y)
dt
dy
Fy (x, y)
dt
The result easily extends to situations where x and y depend on more than one
variable.
We don’t go through a proof, if you are interested, check chapter 12.1.

4 / 36
The chain rule as stated here can be seen as a general statement of all rules of
differentiation from Lecture 1:
Example 2
x
z = F(x, y) = , x = f (t), y = g(t)
y
Application of the chain rule:
 
1 0 x
z (t) = · f (t) + − 2 g0 (t)
0
y y
yf (t) − yg (t) f (t)g(t) − f (t)g0 (t)
0 0 0
= =
y2 [g(t)]2
f (t)
But this is simply applying the quotient rule to the function g(t) .

5 / 36
Linear Approximations

Consider a function f (x, y). For fixed values of x0 , y0 , x, and y, define the
function g(t) by

g(t) = f (x0 + t(x − x0 ), y0 + t(y − y0 ))

In other words, approximating f around (x0 , y0 ) is now equivalent to

approximating g around 0. Note that g(0) = f (x0 , y0 ) and g(1) = f (x, y).

6 / 36
The linear approximation is then a tangent plane:

f (x, y) ≈ f (x0 , y0 ) + fx (x0 , y0 )(x − x0 ) + fy (x0 , y0 )(y − y0 ) (1)

7 / 36
Example 3
Find the linear approximation to f (x, y) = ex+y (xy − 1) about (0, 0). We need:

fx (x, y) = ex+y (xy − 1) + ex+y y

fy (x, y) = ex+y (xy − 1) + ex+y x

So f (0, 0) = −1, fx (0, 0) = −1 and fy (0, 0) = −1. Hence, (1) gives

ex+y (xy − 1) ≈ −1 − x − y

For x and y close to 0, the function z = ex+y (xy − 1) is approximated by the

simple linear function z = −1 − x − y.

8 / 36
The General Case

The linear approximation to z = f (x) = f (x1 , . . . , xn ) about x0 = (x10 , . . . , xn0 ) is

given by

f (x) ≈ f (x0 ) + f1 (x0 )(x1 − x10 ) + · · · + fn (x0 )(xn − xn0 )

9 / 36
Differentials

Suppose that z = f (x, y) is a differentiable function of two variables. Let dx

and dy denote arbitrary real numbers (not necessarily small). Then we define
the differential of z = f (x, y) at (x, y), denoted by dz or df , so that

z = f (x, y) =⇒ dz = fx (x, y) dx + fy (x, y)dy

When x is changed to x + dx and y is changed to y + dy, then the actual change

in the value of the function is the increment

∆z = f (x + dx, y + dy) − f (x, y)

If dx and dy are small in absolute value, then ∆z can be approximated by dz:

∆z ≈ dz = fx (x, y) dx + fy (x, y)dy

10 / 36
11 / 36
Example 4
Let z = f (x, y) = xy. Find dz.

12 / 36
Rules for Differentials

Suppose that f (x, y) and g(x, y) are differentiable, with differentials

df = fx0 dx + fy0 dy and dg = g0x dx + g0y dy.

d(af + bg) = a df + b dg
d(fg) = g df + f dg (2)
 
f g df − f dg
d =
g g2

Chain rule for differentials: Suppose that z = F(x, y) = g(f (x, y)), where g is a
differentiable function of one variable:

z = g(f (x, y)) =⇒ dz = g0 (f (x, y))df (3)

13 / 36
The General Case

The differential of a function z = f (x1 , x2 , . . . , xn ) of n variables is defined as

dz = df = f1 dx1 + f2 dx2 + · · · + fn dxn

The rules for differentials in (2) and (3) are valid for functions of n variables.

14 / 36
Implicit Differentiation along a Level Curve

We will now generalize the implicit differentiation introduced in lectures 3

and 4.
I Let F be a function of two variables, and consider a level curve of F:

F(x, y) = c (c is a constant).

I Suppose this equation defines y implicitly as a function y = f (x) such that

F(x, f (x)) = c for all x in I

15 / 36
Along the level curve, y is an implicit function of x ∈ I.

16 / 36
Example 5
Define F(x, y) = xy.
Then Fx0 (x, y) = y and Fy0 (x, y) = x. Hence

dy Fx (x, y) y
=− =−
dx Fy (x, y) x

17 / 36
The General Case

The method extends to n variables. Consider the function y = F(x), where

x =(x1 , x2 , . . . , xn ). For F(x) = c we have:

Implicit Differentiation
∂F
∂ xi ∂ xj
= − ∂F
∂ xj ∂x i

∂F
given the assumption that ∂ xi 6= 0.

18 / 36
Example 6
Consider the equation x − 2y − 3z + z2 = −2. Find z0x and y0z .

19 / 36
Homogeneous Functions of Two Variables

I If F(K, L) denotes the number of units produced when K units of capital

and L units of labour are used as inputs, what happens to production if
we double the inputs of capital and double the input of labour?
I A function f of two variables x and y defined in a domain D is said to be
homogeneous of degree k if, for all (x, y) in D,

f (tx, ty) = tk f (x, y) for all t > 0 (4)

Multiplication of both variables by a positive factor t will thus multiply

the value of the function by the factor tk .
I The degree of homogeneity of a function can be an arbitrary
number—positive, zero, or negative.

20 / 36
Example 7
The Cobb–Douglas function F defined by F(x, y) = Axa yb is homogeneous of
degree a + b

Example 8
1
f (x, y) = (x−γ + y−γ )− γ is homogeneous of degree 1

21 / 36
Euler’s Theorem

Suppose f (x, y) is homogeneous of degree k. Then we have:

Euler’s Theorem
x fx (x, y) + y fy (x, y) = k f (x, y) (5)

Proof

22 / 36
Example 9
Check (5) for f (x, y) = 3x2 y − y3 .
f (x, y) is homogeneous of degree 3.
fx (x, y) = 6xy and fy (x, y) = 3x2 − 3y2 . Therefore,

xfx (x, y) + yfy (x, y) = x · 6xy + y · (3x2 − 3y2 )

= 3(3x2 y − y3 ) = 3f (x, y)

23 / 36
An often used variant
A version of (4) is:  y
f (x, y) = xk f 1,
x
(Same for y).

Example 10
Suppose a C-D production function Y = K a L1−a . Since

a the function is
homogeneous of degree 1 we can write Y = L · KL . With the capital-labor
ratio κ = KL , we can express efficiency units of labor as

Y
= κa
L

24 / 36
Geometric Aspects of Homogeneous Functions
Let f (x, y) be homogeneous of degree k.

25 / 36
Let f (x, y) be homogeneous of degree k with f (x, y) = c. We can use this level
curve to construct all the other level curves:

26 / 36
General Homogeneous Functions

Suppose that f is a function of n variables defined in a domain D. Suppose

that whenever (x1 , x2 , . . . , xn ) ∈ D and t > 0, then (tx1 , tx2 , . . . , txn ) also lies in
D. (A set D with this property is called a cone.) We say that f is
homogeneous of degree k on D if

f (tx1 , tx2 , . . . , txn ) = tk f (x1 , x2 , . . . , xn ) for all t > 0

The constant k can be any number—positive, zero, or negative.

27 / 36
Euler’s theorem generalizes. In particular, for f homogeneous of degree k we
have:
n
∑ xi fi0 (x) = kf (x)
i=1

Dividing the equality by f (x):

n n
xi 0
∑ εif (x) = ∑ fi (x) = k
i=1 i=1 f (x)

28 / 36
Example 11
Consider the general Cobb–Douglas production function
F(v1 , . . . , vn ) = Ava11 · · · vann .

F(tv) = A(tv1 )a1 . . . (tvn )an = Ata1 va11 . . . tan vann = ta1 +···+an F(v)

So F is homogeneous of degree a1 + · · · + an .
I Economists call this the “returns to scale”
I If a1 + · · · + an = 1: constant returns to scale
I If a1 + · · · + an > 1: increasing returns to scale
I If a1 + · · · + an < 1: decreasing returns to scale
I Because εiF = ai , i = 1, . . . , n, we get ∑ni=1 εiF = ∑ni=1 ai

29 / 36
Systems of Equations
I Many economic models relate a large number of variables to each other
through a system of simultaneous equations.
I Let x1 , x2 , . . . , xn be n variables.
I There are n degrees of freedom, because all n variables can be freely
chosen.
I Whenever one “independent” restriction is introduced, the number of
degrees of freedom is reduced by 1.
I In general, introducing m < n independent restrictions on the variables
x1 , x2 , . . . , xn means that the equilibrium variables must satisfy a system of
m independent equations having the form
 1

 f (x1 , x2 , · · · , xn ) = 0
 2
f (x1 , x2 , · · · , xn ) = 0

 ..............................
 m
f (x1 , x2 , · · · , xn ) = 0

30 / 36
The Counting Rule
If n > m, there are n − m degrees of freedom in the system. If n < m, there is
in general no solution to the system.

Note
The counting rule is not generally valid. For example, if 100 variables
x1 , . . . , x100 are restricted to satisfy one equation, the rule says that the number
of degrees of freedom should be 99. However, if the equation happens to be

x12 + x22 + · · · + x100

2
=0

then there is only one solution, x1 = x2 = · · · = x100 = 0, so there are no

degrees of freedom.

31 / 36
Degrees of Freedom
A system of equations in n variables is said to have k degrees of freedom if
there is a set of k variables that can be freely chosen, while the remaining
n − k variables are uniquely determined once the k free variables have been
assigned specific values.
A system with m = n is usually consistent (that is, has solutions), but it may
have several solutions. A system of equations does not, in general, have a
unique solution unless there are exactly as many equations as unknowns.

32 / 36
Example 12
Consider the macroeconomic model
(i) Y = C+I +G
(ii) C = f (Y − T)
(iii) I = h(r)
(iv) r = m(M)

where f , h, and m are given functions.

Y: national income, C: consumption, I: investment, G: public expenditure, T:
tax revenue, r: interest rate, M: money supply
How many degrees of freedom are there?
The system determines the variables Y, C, I, and r as functions of M, T, and
G.

33 / 36
The General Case

We distinguish between:
I endogenous variables
I exogenous variables

34 / 36
A general system of structural equations:
 1

 f (x1 , x2 , ..., xn , y1 , y2 , ..., ym ) = 0
 2
f (x1 , x2 , ..., xn , y1 , y2 , ..., ym ) = 0
(6)

 ..............................
 m
f (x1 , x2 , ..., xn , y1 , y2 , ..., ym ) = 0

I System (6) with m equations in n + m unknowns has n + m − m = n

degrees of freedom. Suppose it defines all the endogenous variables
y1 , . . . , ym as C1 functions of x1 , . . . , xn . Then the system can be solved “in
principle” for y1 , . . . , ym in terms of x1 , . . . , xn to give the reduced form:

y1 = ϕ 1 (x1 , . . . , xn )
...
ym = ϕ m (x1 , . . . , xn )

35 / 36
Example 13
Suppose we want to model the price in the market for used cars. The
structural equations in this model are:

D = f (p, q, m, t) and S = g(p, r)

where

p = price of used cars

q = price of new cars
m = average income of the population
t = price of public transport
r = market price for used cars overseas

The solution will be a reduced form equation, which gives the endogenous
equilibrium price as a function of the exogenous variables:

p∗ = F(q, m, t, r)
36 / 36
Multivariate Optimization
Mathematics I
Lectures 13-14

Dr. Clemens Buchen

1 / 37
Plan of the lecture

1. Necessary conditions
2. Sufficient conditions
3. Extreme value theorem
4. Envelope theorem
5. Many variables - some generalizations
Based on chapter 13

2 / 37
I Most interesting economic optimization problems require the
simultaneous choice of several variables.
I For example, a profit-maximizing producer of a single commodity
chooses quantities of many different inputs, as well as its output level. A
consumer chooses what quantities of the many different goods to buy,
etc.
I Most of the difficulties resulting from multivariate optimization already
arise in the transition from one to two variables so that most of the class
will focus on this case.

3 / 37
Maximum

4 / 37
Two Variables: Necessary Conditions
I Consider a function z = f (x, y) defined on a set S in the xy-plane.
Suppose that f attains its maximum at an interior point (x0 , y0 ) of S.
I If we keep y fixed at y0 , then the function g(x) = f (x, y0 ) depends only on
x and has its maximum at x = x0 .
I g(·) is a function of one variable. We know that the first-order condition
g0 (x0 ) = 0 must be satisfied.
I Hence fx (x0 , y0 ) = 0 is necessary, since

g0 (x0 ) = 0 ⇔ fx (x0 , y0 ) = 0

I Applying the same logic w.r.t y, we conclude that fy (x0 , y0 ) = 0 is

necessary.
I If f attains its smallest value (its minimum) at an interior point (x0 , y0 ) of
S, a similar argument holds.
I A point (x0 , y0 ) where both the partial derivatives are 0 is called a
critical point or stationary point of f .
Necessary Conditions for Interior Extrema

Theorem
A differentiable function z = f (x, y) can only have a maximum or minimum at
an interior point (x0 , y0 ) of S if it is a critical point—that is, if the point
(x, y) = (x0 , y0 ) satisfies the two equations

fx (x0 , y0 ) = 0
(first-order conditions, or FOCs).
fy (x0 , y0 ) = 0

6 / 37
Example 1
Consider the function f : R2 7−→ R

f (x, y) = −2x2 − 2xy − 2y2 + 36x + 42y − 158

Assume that f has a maximum point. According to the foregoing theorem, at

the maximum the first-order conditions must hold:

fx (x, y) = −4x − 2y + 36 = 0,
fy (x, y) = −2x − 4y + 42 = 0.

The linear equation system has a unique solution (x, y) = (5, 8). If it really is a
maximum, f takes a maximal value of f (5, 8) = 100.

7 / 37
Example 2
Suppose that Q = F(K, L) is a production function with
K = capital input with r = the cost per unit of capital
L = labour input with w = the wage rate
Q = output with p = output price
Denoting the profit with π and assuming that all production can be sold at the
market price, the firm’s profit becomes:

π(K, L) = pF(K, L) − rK − wL

8 / 37
Sufficient Conditions for a Maximum and Minimum
Suppose that (x0 , y0 ) is an interior critical point for a C2 -function f (x, y) in a
convex set S.
(a)
If for all (x, y) in S,

fxx (x, y) ≤ 0, fyy (x, y) ≤ 0, and fxx (x, y)fyy (x, y) − (fxy (x, y))2 ≥ 0

then (x0 , y0 ) is a maximum point for f (x, y) in S. The function is concave.

(b)
If for all (x, y) in S,

fxx (x, y) ≥ 0, fyy (x, y) ≥ 0, and fxx (x, y)fyy (x, y) − (fxy (x, y))2 ≥ 0

then (x0 , y0 ) is a minimum point for f (x, y) in S. The function is convex.

9 / 37
Example 3
Consider the earlier example:

f (x, y) = −2x2 − 2xy − 2y2 + 36x + 42y − 158.

The first-order conditions are

fx (x, y) = −4x − 2y + 36,

fy (x, y) = −2x − 4y + 42.

Setting the FOCs equal to zero and solving yields (x0 , y0 ) = (5, 8).
The second order partials are fxx (x, y) = −4, fyy (x, y) = −4 and
fxy (x, y) = −2. Thus

fxx (x, y) = −4 ≤ 0, fyy (x, y) = −4 ≤ 0 and fxy (x, y) = −2

fxx (x, y)fyy (x, y) − (fxy (x, y))2 = 16 − 4 = 12 ≥ 0.

Thus the stationary point (5, 8) is a maximum point.

10 / 37
Local Extreme Points

11 / 37
I The point (x0 , y0 ) is said to be a local maximum point of f in S if
f (x, y) ≤ f (x0 , y0 ) for all pairs (x, y) in S that lie sufficiently close to
(x0 , y0 ).
I The point (x0 , y0 ) is said to be a local minimum point of f in S if
f (x, y) ≥ f (x0 , y0 ) for all pairs (x, y) in S that lie sufficiently close to
(x0 , y0 ).
I The point (x0 , y0 ) is said to be a saddle point of f in S if
f (x, y) ≥ f (x0 , y0 ) for some pairs (x, y) in S that lie sufficiently close to
(x0 , y0 ) and also f (x, y) ≤ f (x0 , y0 ) for some pairs that lie sufficiently
close to (x0 , y0 )

12 / 37
Second-Derivative Test for Local Extrema

Suppose f (x, y) is a function with continuous second-order partials in S.

Suppose (x0 , y0 ) is a critical point in the interior of S.

A = fxx (x0 , y0 ), B = fxy (x0 , y0 ), and C = fyy (x0 , y0 )

(i) If A, C < 0 and AC − B2 > 0, then (x0 , y0 ) is a (strict) local maximum

point.
(ii) If A, C > 0 and AC − B2 > 0, then (x0 , y0 ) is a (strict) local minimum
point.
(iii) If AC − B2 < 0, then (x0 , y0 ) is a saddle point.
(iv) If AC − B2 = 0, then (x0 , y0 ) could be a local maximum, a local
minimum, or a saddle point.
We omit the proof (see page 507 in the textbook)
Example 4
Consider the function f (x, y) = x3 − x2 − y2 + 8. (a) Find the stationary points
and (b) classify them.
(a) Stationary points: Set the two first-order partials equal to zero:

3x2 − 2x = x(3x − 2) = 0
−2y = 0

This implies two points: (0, 0) and ( 32 , 0).

14 / 37
(b)
Find all second-order partials: fxx = 6x − 2, fyy = −2, fxy = 0
Make a table (A, B, C as defined above):

A B C AC − B2
(0, 0) −2 0 −2 4
( 23 , 0) 2 0 −2 −4

Therefore (0, 0) is a local maximum and ( 23 , 0) is a saddle point.

15 / 37
A set S ⊆ R2

Some definitions:
I A point (a, b) ∈ S is called an interior point if there exists a circle
centred at (a, b) such that all points within the circle lie in S.
I A set is called open if it consists only of interior points (it generalizes the
concept of an open interval).
I The point (a, b) is called a boundary point of a set S if every circle
centred at (a, b) contains points of S as well as points in its complement.
I If S contains all its boundary points, then S is called closed (it
generalizes the concept of a closed interval). A set is closed if and only if
its complement is open.

16 / 37
17 / 37
Example 5
Provided that p, q, and m are positive parameters, the budget set is a set of
points (x, y) that satisfy the inequalities

px + qy ≤ m, x ≥ 0, y≥0

is a closed set.

B(p, q, m) = (x, y) ∈ R2 | px + qy ≤ m, x ≥ 0, y ≥ 0


The set that results from replacing ≤ by < and ≥ by > is open.

18 / 37
In general, if g(x, y) is a continuous function and c is a real number, then the
sets

A = {(x, y) : g(x, y) ≥ c },
B = {(x, y) : g(x, y) ≤ c },
C = {(x, y) : g(x, y) = c}

are all closed.

If ≥ is replaced by >, or ≤ is replaced by <, or = by 6=, then the
corresponding set becomes open.

19 / 37
Compact sets

A set in the plane is bounded if the whole set is contained within a

sufficiently large circle.
Definition
A set in the plane that is both closed and bounded is called compact.

Example 6
Construct a set in R2 , which is closed, but not bounded.

20 / 37
Extreme Value Theorem

Suppose the function f (x, y) is continuous throughout a nonempty and

compact set S ⊆ R2 . Then there exists both a point (a, b) in S where f has a
minimum and a point (c, d) in S where f has a maximum – that is,

f (a, b) ≤ f (x, y) ≤ f (c, d) for all (x, y) ∈ S.

21 / 37
Finding Maxima and Minima

Finding the maximum and minimum values of a differentiable function f (x, y)

defined on a compact set S in the plane:
(I) Find all stationary points of f in the interior of S.
(II) Find the largest value and the smallest value of f on the boundary of S
(or each of its subdivisions), along with the associated points.
(III) Compute the values of the function at all the points found in (I) and (II).
The largest function value is the maximum value of f in S. The smallest
function value is the minimum value of f in S.

22 / 37
Example 7
Find the extreme values for f (x, y) defined over S when

f (x, y) = x2 + y2 + y − 1, S = {(x, y) : x2 + y2 ≤ 1 }

(I) The stationary points in the interior of S satisfy the two equations

fx (x, y) = 2x = 0, fy (x, y) = 2y + 1 = 0

(x, y) = (0, − 21 ) is the only stationary point. Moreover it is in the interior

of S, with f (0, − 12 ) = −5/4.
(II) The boundary of S consists of the circle x2 + y2 = 1
(III) Minimum at (x, y) = (0, − 21 ), maximum at (x, y) = (0, 1)

23 / 37
24 / 37
Envelope Theorem
Consider the problem
max f (x, r).
x

The value of x that maximizes f will usually depend on r, so we denote it by

x∗ (r). Implicitly x∗ (r) is defined by the first-order condition:

fx (x∗ (r), r) = 0.

Inserting x∗ (r) into f (x, r), we obtain the value function:

v∗ (r) = f (x∗ (r), r)

Alternatively, we can write:

v∗ (r) = max f (x, r).

x

25 / 37
 v∗ (r) = max f (x, r).
x

What happens to the value function as r changes? Assuming that v∗ (r) is

differentiable:
dv∗ dx∗
(r) = fx (x∗ (r), r) (r) + fr (x∗ (r), r).
dr | {z } dr
=0 due to foc

Thus, we obtain:
dv∗
(r) = fr (x∗ (r), r).
dr
This result is called the envelope theorem.

26 / 37
Example 8
Consider a firm that produces at the cost C(x) = x2 and sells at the constant
price r. The firm’s maximization problem is

π ∗ (r) = max π(x, r) = rx − x2

x

where π(x, r) denotes profit and π ∗ (r) the value function.

First-order condition:

r − 2x∗ = 0 ⇐⇒ x∗ (r) = r/2

27 / 37
The value function, i.e. maximum profit, is

π ∗ (r) = rx∗ (r) − [x∗ (r)]2

r  r 2 r2
= r − = .
2 2 4
It implies
dπ ∗ r
(r) = 2 = r/2.
dr 4
Applying the envelope theorem yields the result directly:

dπ ∗
(r) = πr (x∗ (r), r) = x∗ (r) = r/2.
dr

28 / 37
Numerical example:
5

3
x=1
x=2
2 x=x * (r)

r
0 1 2 3 4 5

29 / 37
Generalization of the Envelope Theorem

Let x = (x1 , . . . , xn ), r = (r1 , . . . , rk ) and define the value function f ∗

f ∗ (r) = max f (x, r)

x

and if x∗ (r) is the value of x which maximizes f (x, r), then

∂f∗ ∂f ∗
(r) = (x (r), r) for all j = 1, . . . , k
∂ rj ∂ rj

provided that the partial derivative exists.

30 / 37
n variables

In the following we generalize the results for functions with n variables.

1. Definition of an extreme point
2. Necessary conditions
3. Sufficient conditions
4. Sets
5. Extreme value theorem

31 / 37
1. Definition of an extreme point
If f (x) = f (x1 , . . . , xn ) is a function of n variables defined over a set S in Rn ,
then c = (c1 , . . . , cn ) is a (global) maximum point for f in S if

f (x) ≤ f (c) for all x in S

Similar for a minimum.

32 / 37
2. Necessary conditions
Suppose f is defined in a set S in Rn and let c = (c1 , . . . , cn ) be an interior
point in S where f is differentiable. A necessary condition for c to be a
maximum or minimum point for f is that c is a stationary point for f i.e.

fi (c) = 0, i = 1, . . . , n

33 / 37
3. Sufficient conditions
Suppose a function z = f (x) where x = (x1 , . . . , xn )
Recall the Hessian:
 
f11 (x) f12 (x) ... f1n (x)
 f21 (x) f22 (x) ... f2n (x) 
D2 f (x) = f 00 (x) = 
 :

: : 
fn1 (x) fn2 (x) ... fnn (x)

The function is concave (convex) iff D2 f (x) is negative (positive) semidefinite.

Then the stationary point c = (c1 , . . . , cn ) is a global maximum (minimum).
Definiteness of a matrix is found by checking the determinants of D2 f (x).
We’ll return to that.

34 / 37
4. Sets
We need some additional definitions:
Define the distance between the points x = (x1 , . . . , xn ) and y = (y1 , . . . , yn ) in
Rn by q
kx − yk = (x1 − y1 )2 + (x2 − y2 )2 + · · · + (xn − yn )2
If y = 0, then q
kxk = x12 + x22 + · · · + xn2
is the distance between x and the origin. The number kxk is also called the
norm of the vector x.

35 / 37
An open (n-dimensional) ball with centre at a = (a1 , . . . , an ) and radius r is
the set of all points x = (x1 , . . . , xn ) such that kx − ak < r.

The definitions in the plane (i.e. R2 ) of interior point, open set, boundary
point, closed set, bounded set, and compact set all generalize for sets in Rn if
we replace the word “circle” by “ball”.

If g(x) = g(x1 , . . . , xn ) is a continuous function, and c is a real number, then

each of the three sets

{x : g(x) ≥ c}, {x : g(x) ≤ c }, {x : g(x) = c}

is closed. If ≥ is replaced by >, ≤ by <, or = by 6=, the corresponding set is

open.

36 / 37
5. Extreme Value Theorem
Suppose the function f is continuous throughout a nonempty, closed and
bounded set S ⊂ Rn . Then there exist both a point d ∈ S where f has a
minimum and a point c ∈ S where f has a maximum – that is,

f (d) ≤ f (x) ≤ f (c) for all x in S

37 / 37