Mathematics for economists

Université Paris-Dauphine
Bachelor in Economics
R. Aı̈d, H. Malamut, E. Ndiaye, M. Oury and L. Veraldi∗
2024-2025

∗ Jérôme Renault and Tristan Tomala (2006-2007), Olivier Rivière (2007-2008), Cyril Im-

bert (2007-2010), Anne-Marie Boussion (2011-2014), Vincent Iehlé (2009-2014) and Françoise
Forges (2016-2019) also largely contributed to writing these notes.

1
Introduction
This course focuses on a recurring problem in economics: optimizing an
objective function under constraints. It assumes some familiarity with basic
mathematical concepts such as natural integers (set N), real numbers (set R),
functions of a real variable (limits, continuity, derivative,...) but briefly recalls
most useful definitions. These notes should be understood as a summary
of the course, but additional comments, details and examples will be
given orally.
The reader may also consult:
Philippe Michel’s book: “Cours de mathématiques pour économistes”, Edi-
tions Economica, 1989.
Martin Osborne’s online course “Mathematical methods for economic the-
ory”, freely accessible on the author’s Internet site:
https://mjo.osborne.economics.utoronto.ca/index.php/tutorial/index/1/int/i

Table of content
1. Some basic optimization notions
1.1 Extremum of a subset of R and of a function
1.2 Elements of topology of Rd
1.2.1 Norms and distances on Rd
1.2.2 Convergent sequences, continuous functions
2. First notions of optimization in Rd
2.1 Notions of differential calculus
2.2 First and second order conditions
2.2.1 First order necessary conditions
2.2.2 Second order conditions
2.3 Concave case
2.3.1 Convex sets
2.3.2 Concave functions
2.4 Functions defined on a compact set, coercive functions
3. Constrained optimization in Rd
3.1 Karush-Kuhn-Tucker conditions.
3.2 Optimization with equality constraints
3.2.1 A single equality constraint
3.2.2 Several equality constraints
3.3 Optimization with inequality constraints
3.4 Optimization with equality and inequality constraints

2
1 Basic notions for optimization problems
The aim of this course is to rigorously describe an optimization problem of type:

max f (x) (also denoted by: max {f (x) : x ∈ C} ) (1)

x∈C

and to understand methods that allow solving it. The aim is to choose a variable
x (which, typically, will have several coordinates) in order to obtain the best
possible result f (x), where f is an “objective function”, f : C → R, defined on
a set C describing “contraints”.
A well-known example in economics is the example of a consumer who needs
to choose a quantity x1 of good 1, and x2 of good 2, in order to maximize its
utility f (x1 , x2 ), which represents its preferences, under a budget constraint,
that depends on its income R and prices (p1 > 0 for good 1, p2 > 0 for good 2):

C = (x1 , x2 ) ∈ R2 : x1 ≥ 0, x2 ≥ 0, p1 x1 + p2 x2 ≤ R ⊆ R2 .

(2)

In practice, there is no reason to limit the number of coordinates of decision

variable x. In the previous example, there could be more than two goods. The
set C describing constraints will thus be a subset of Rd , where d is a positive
integer (in our example, the number of goods). An element - or “vector” - x
of Rd is described by d coordinates xi ∈ R (or sometimes xi ), i = 1, ..., d. It is
denoted by x = (xi )1≤i≤d , that is,
 
x1
 .. 
 . 
 
x= xi  .

 . 
 .. 
xd

For instance, when d goods are available, the budget set of a consumer with
income R is
( d
)
X
d
C = x ∈ R : xi ≥ 0, i = 1, ..., d, p i x i ≤ R ⊆ Rd , (3)
i=1

where pi > 0 is the price of good i (i = 1, ..., d).

A first question is whether Problem (??) has a solution, that is, if there
exists a choice x∗ (at least one) satisfying all constraints (x∗ ∈ C) and that is
the best possible, given these constraints:

∀x ∈ C : f (x) ≤ f (x∗ ).

Below (Section 1.1), the potential solutions of problem (??) will be characterized
in details by considering the image of set C by f , that is

f (C) = {f (x) : x ∈ C} = {y ∈ R : ∃ x ∈ C such that y = f (x)} , (4)

3
and by checking whether set f (C) ⊆ R has a maximum. Thus, as we will see,
the first step to study Problem (??) is to understand the notion of maximum of
a set A ⊆ R, keeping in mind the case A = f (C).
Later (Section 1.2), in order to identify useful properties of the constraint
set (the set C of points satisfying the constraints) of Problem (??), topologi-
cal properties of set Rd will be studied. A fundamental property of objective
function f , continuity, will also be defined and characterized.
More precise properties of the constraint set and of the objective function
will be studied in Chapter 2. This will allow us to define conditions for Problem
(??) to have a solution.

1.1 Extremum of a subset of R and of a real-valued func-

tion
This section groups useful definitions to study the extrema (that is, the maxi-
mum or the minimum) of a set of real numbers.

Definition (Maximum and minimum) Let A be a nonempty set of real numbers

(A ⊆ R) and let a ∈ R.

- a is the maximum of A if a ∈ A and for all b ∈ A, b ≤ a. In that case, we

write a = max A. We also say that a is the greatest element of A.
- a is the minimum of A if a ∈ A and for all b ∈ A, b ≥ a. In that case, we
write a = min A. We also say that a is the least element of A.

Proposition The maximum and the minimum of a set A ⊆ R, if they exist,

are unique.

Proof Let a and a0 ∈ R. Assume that a and a0 both satisfy the definition of the
maximum of A. By the first part of the definition, a ∈ A and a0 ∈ A. Therefore,
by the second part applied to a (with b = a0 ), we obtain a0 ≤ a. Exchanging the
roles of a and a0 , we obtain a ≤ a0 . Therefore a = a0 .

Exercise 1.1 Find, if they exist, the maximum √ and the minimum of the
following sets: [0, 1[, ]2, 3], ]4, 5[, ] − ∞, π] and ] 2, +∞[.

Definition (Set bounded from above, bounded from below, bounded) Let A be
a nonempty set of real numbers (A ⊆ R).

- Number a ∈ R is an upper-bound of A if for all b ∈ A, b ≤ a.

- Number a ∈ R is a lower-bound of A if for all b ∈ A, b ≥ a.
- Set A is bounded from above if it has an upper-bound, that is, if there exists
M ∈ R such that for all a ∈ A, a ≤ M .

4
- Set A is bounded from below if it has a lower-bound, that is, if there exists
m ∈ R such that for all a ∈ A, a ≥ m.
- Set A is bounded if it is bounded from above and from below, that is, if there
exists m ∈ R and M ∈ R such that for all a ∈ A, m ≤ a ≤ M .

The previous exercise shows that a maximum and a minimum need not
exist.1 This motivates the introduction of the concept of least upper-bound (or
supremum) and greatest lower-bound (or infimum).

Definition (least upper-bound and greatest lower-bound) Let A be a nonempty

set of real numbers (A ⊆ R).

- Number a ∈ R is the least upper-bound of set A if A has an upper-bound

and a is the smallest one. That is, a is the minimum of the set of upper-
bounds of A. We also say that a is the supremum of A. It is denoted by
a = sup A.
- Number a ∈ R is the greatest lower-bound of set A if A has a lower-bound and
a is the greatest one. That is, a is the maximum of the set of lower-bounds
of A. We also say that a is the infimum of A. It is denoted by a = inf A.

Proposition The least upper-bound and the greatest lower-bound of a set

A ⊆ R, if they exist, are unique.

Exercise 1.2 Prove the previous proposition.

We admit the following fundamental result

Proposition In R, any nonempty set that is bounded from above has a least
upper-bound (in R); any nonempty set that is bounded from below has a greatest
lower-bound (in R).

Exercise 1.3 Consider the sets of Exercise 1.1. Check whether the upper-
bounds, lower-bounds, least upper-bounds, and greatest lower-bounds of these
sets exist, and in this case, find them.

Proposition Let A ⊆ R and a ∈ R. The following properties are equivalent:

1. a = sup A
2. For all b in A, b ≤ a; moreover, for all c in R, if c < a, then c is not an
upper-bound of A.
1 Note to French students: this means that they do not necessarily exist. They may exist,

but this is not always the case.

5
 3. For all b ∈ A, b ≤ a; moreover, for all ε > 0, there exists aε ∈ A such that
aε > a − ε.

Remark: the piece of notation aε above aims at clarifying that the element aε
depends on ε.

Exercise 1.4 Write the analog of the previous proposition for the infimum.

The next proposition relates the notions of maximum and supremum, and
the notions of minimum and infimum.

Proposition Let A ⊆ R and a ∈ R.

- If the least upper-bound sup A of A exists and belongs to A (that is, sup A ∈
A), then A has a maximum and max A = sup A. Similarly, if the greatest
lower-bound inf A of A exists and inf A ∈ A, then A has a minimum and
min A = inf A.
- If A has a maximum, then A has a supremum and sup A = max A. Similarly,
if A has a minimum, then A has an infimum and inf A = min A.

Exercise 1.5 Prove the previous proposition.

Here are common conventions, that we will use:

- if set A is not bounded from above, we write sup A = +∞;

- if set A is not bounded from below, we write inf A = −∞;
- if set A is empty (A = ∅), we write sup A = −∞ and inf A = +∞.

Let us now consider, as in the beginning of Chapter 1, a real-valued function

f : C → R, defined on an arbitrary set C (in the course, C will be a subset of
Rd ). The image of C by f is denoted by f (C) and defined by Eq. (??).

Definition (Extrema of a real-valued function) Let f : C → R.

- The supremum and the infimum of f (on C) are the supremum and the
infimum of set f (C). They are denoted by sup f (C), inf f (C), or by
supC f , inf C f , or by sup{f (x) : x ∈ C}, inf{f (x) : x ∈ C}. Similarly, if
they exist, the maximum and the minimum of f (on C) are the maximum
and the minimum of set f (C). They are denoted by max f (C), min f (C),
or by maxC f , minC f , or by max {f (x) : x ∈ C}, min {f (x) : x ∈ C}.
- Function f is bounded from above (on C) if set f (C) is bounded from above.
Similarly, f is bounded from below (on C) if set f (C) is bounded from
below.

6
 Let us come back to the optimization problem maxx∈C f (x) considered at
the beginning of Chapter 1: function f described the objective, and set C the
constraints.

Definition (Value and solution) Let us consider the optimization problem (??).

- The value of Problem (??) is the number supC f .

- The solutions of Problem (??) are the elements x∗ ∈ C such that supC f =
f (x∗ ). If solutions exist, then supC f = maxC f .

Remark: a similar definition applies to the optimization problem minx∈C f (x).

Example: let C = [0, 1[ and f (x) = 2x + 1. The value of the minimization
problem minx∈C f (x) is 1 and x∗ = 0 is the solution of this problem. By contrast,
the maximization problem maxx∈C f (x) has a value, supx∈C f (x) = 3, but no
solution.
Exercise 1.6 Compute the suprema of the following functions on the indicated
sets:

• 2x on C =] − 1, 1[
1
• 1− x on C = [1, +∞[
• sin(x) on C = R.
• f (x) = 2x + 1 if 0 < x < 1, f (0) = 0, f (1) = 2 on C = [0, 1].

To do so, it may be helpful to draw the graph of these functions. Are the
suprema maxima? What are the values of these maximization problems? Do
these problems have solutions?

1.2 Elements of topology of Rd

1.2.1 Norms and distances on Rd
To measure the distance between two elements x and y of Rd , it is useful to
introduce the notion of norm on Rd : a generalization of the notion of absolute
value on R.

Definition (Norm) A norm on Rd is a mapping k · k: Rd → R+ , (hence with

nonnegative values, that is, such that for all x in Rd , k x k≥ 0), satisfying the
following properties:

1. ∀ x ∈ Rd , k x k= 0 ⇔ x = 0
2. ∀ x, y ∈ Rd , k x + y k≤k x k + k y k
3. ∀ x ∈ Rd , ∀λ ∈ R, k λx k= |λ| k x k.

7
Exercise 1.7 Represent graphically the absolute value function | · |: R → R+
and show that this is indeed a norm on R.

Exercise 1.8 Check that: ∀ x, y ∈ Rd , k y − x k=k x − y k.

Exercise 1.9 ? Check that: ∀ x, y ∈ Rd , |k x k − k y k|≤k x − y k.

Definition (Distance associated to a norm) Let k · k be a norm on Rd .

For all x, y in Rd , we denote by d(x, y) =k x − y k the distance between x and
y associated to this norm.

Exercise 1.10 Check that the above defined distance satisfies the following
properties:

1. ∀ x, y ∈ Rd , d(x, y) = 0 ⇔ x = y.
2. ∀ x, y ∈ Rd , d(x, y) = d(y, x).
3. ∀ x, y, z ∈ Rd , d(x, y) ≤ d(x, z) + d(z, y).

Definition (Standard norms on Rd )

s
d
P
k x k2 = (xi )2 (euclidean norm);
i=1

d
P
k x k1 = |xi | (norm 1)
i=1

k x k∞ = max1≤i≤d {|xi |} (sup norm or infinite norm).

Remark (out of program) : For each real number p ≥ 1, it is possible to define

norm p, by the following formula:

d
!1/p
X
p
k x kp = (xi )
i=1

The 1 and 2- norms are obtained by taking p = 1, p = 2. The infinity norm

(also called sup norm) is obtained by making p tend toward infinity.

d
Definition (Scalar products on Rd ) ∀ x, y ∈ Rd , x · y =
P
xi yi denotes the
i=1
d
2
scalar product of x and y. (In particular: ∀ x ∈ Rd , x · x = x2i = (k x k2 ) ).
P
i=1

Exercise 1.11 ?? Prove Cauchy-Schwartz inequality, that is:

∀x, y ∈ Rd , |x · y| ≤k x k2 k y k2

8
 2
Hint: fix arbitrarily x, y ∈ Rd . Note that (k x + ay k2 ) ≥ 0 for all a in R. Thus,
considering the left-hand-side as a second order polynomial in a, its discriminant
must be nonpositive.

Exercise 1.12 ? Check that the standard norms on Rd are indeed norms, in
the sense of the above definition. Hint: for the euclidean norm, use Cauchy-
Schwartz inequality (see previous exercise).

Definition-Proposition (Equivalence of norms) Two arbitrary norms on Rd ,

denoted by k · k and k · k0 , are equivalent if there exist constants a > 0 and
b > 0 such that, for all x ∈ Rd ,

a k x k≤k x k0 ≤ b k x k .

The above inequalities are satisfied if and only if there exist constants a0 > 0
and b0 > 0 such that, for all x ∈ Rd ,

a0 k x k0 ≤k x k≤ b0 k x k0 .

The roles of the two norms are thus symmetrical.

Exercise 1.13 ? Show that if two norms on Rd , denoted by k · k and k · k0 ,

are equivalent with a third one k · k00 then they are equivalent.

Exercise 1.14 Show that the three standard norms on Rd are equivalent.

We admit the following more general result:

Theorem All norms on Rd are equivalent.

From now on, k · k denotes an arbitrary norm on Rd . Though we will

not systematically prove it, the main notions that we will define (continuity,
differentiability, interior of a set,...) do not depend on the chosen norm.

Definition (Open and closed balls) The open ball with center x ∈ Rd and
radius R > 0 is the set

B(x, R) = y ∈ Rd : k x − y k< R .


Similarly, the closed ball with center x and radius R is the set

B(x, R) = y ∈ Rd : k x − y k≤ R .


Exercise 1.15 Check that in R equipped with the absolute value, B(x, R) =
]x − R, x + R[. Give a similar result for the closed ball.

Exercise 1.16 Draw the closed ball B(0, 1) for the three standard norms on
R2 .

9
 The following definitions (point in the interior of a set A, bounded set) will
play a key role in following chapters. Set A will then represent the set of points
satisfying the constraints of an optimization problem.

Definition (Interior point) Let A ⊆ Rd ; x is in the interior of A if there exists

a number ε > 0 such that B(x, ε) ⊆ A. The interior of A (that is, the set of
points in the interior of A) is denoted by intA.

If x is in the interior of A, x ∈ A, and moreover, x may be surrounded by a

ball B(x, ε), whose radius may be very small, but that is entirely in A.

Exercise 1.17 Let I be a bounded interval of R. I may be of the form [a, b],
[a, b[, ]a, b] or ]a, b[, with a, b ∈ R, a < b. Show that x is in the interior of I if
and only if x ∈ ]a, b[ (so that, in any case, int I = ]a, b[).

Exercise 1.18 ?? Show that x is in the interior of the closed ball B(0, 1) (for
an arbitrary norm on R2 ) if and only if x ∈ B(0, 1), the corresponding closed
ball (so that intB(0, 1) = B(0, 1)).

Exercise 1.19 Show that the interior of Rd is Rd itself.

Definition (Bounded set) Set A ⊆ Rd is bounded if there exists a number

M > 0 such that for all x in A, k x k≤ M .

Remark: the order of the quantifiers is crucial. It indicates that the same
number M works for all elements of A.

Exercise 1.20 Let A and B be two subsets of Rd . Show that if B is bounded

and A ⊆ B, A is also bounded.

Exercise 1.21 Show that any ball (whether open or closed) is bounded.

Exercise 1.22 Show that a set A ⊆ Rd is bounded if and only if it is a subset

of a ball (not necessarily centered on 0). That is, A ⊆ Rd is bounded ⇔ ∃
x ∈ Rd and R > 0 such that A ⊆ B(x, R).

Exercise 1.23 Check that if A is a set of real numbers (A ⊆ R), the above
definition is equivalent to the definition of Section 1.1. That is, A is bounded if
and only if A is bounded from above and from below.

Exercise
 1.24 Are the following sets bounded?
A = (x, y) ∈ R2 : x2 + y 2 ≤ 1 ,
C = (x, y) ∈ R2 : x2 − y 2 ≤ 1 ,
D = (x, y) ∈ R2 : y ≤ x2 .

10
1.2.2 Convergent sequences, continuous functions
A sequence in Rd is an infinite list x(n), n = 0, 1, 2, ... of elements of Rd (thus,
each of these element has d coordinates: x(n) = (xi (n))1≤i≤d ). Formally, a
sequence in Rd is a mapping: N → Rd : n → x(n). To avoid cumbersome nota-
tion, a sequence in Rd is usually denoted by xn , instead of x(n). Similarly, for
1 ≤ i ≤ d, the ith coordinate of x(n) will be denoted by xni or xin instead of
xi (n).

Definition(Suite convergente) A sequence (xn )n≥0 in Rd converges towards

x ∈ Rd if
lim k xn − x k= 0
n→∞

In this definition, the sequence dn =k xn − x k= d(xn , x) that appears is a

sequence of real numbers, to which we may apply the definition of the limit of
a sequence in R. Doing so, we may rewrite the previous definition as follows:

∀ε > 0, ∃n0 ∈ N, ∀n ≥ n0 , k xn − x k≤ ε (5)

In this expression, the order of the quantifiers is important: n0 is chosen as a

function of ε, but then, for any n ≥ n0 , the distance between xn and x must
be less than ε.

Proposition The limit of a sequence in Rd , if it exists, is unique.

Exercise 1.25 Prove the previous proposition.

Exercise 1.26 ? Let (xn )n≥0 be a sequence in Rd . Show that if (xn )n≥0
converges towards x ∈ Rd , then k xn k converges towards k x k (in R). Hint:
use the result of Exercice 1.9.

Exercise 1.27 Check that if two norms on Rd , k · k and k · k0 , are équivalent,

then any sequence that converges for one of these norms, converges for the
other, and with the same limit. (This is why we did not specify the norm in
the definition of a convergent sequence).

Proposition A sequence (xn )n≥0 in Rd converges towards x ∈ Rd ⇔ ∀ i =

1, ..., d, the sequence (xin )n≥0 (in R) converges towards xi ∈ R.

Exercise 1.28 ?? Prove the previous proposition. Hint: choose cleverly the
underlying norm.

Exercise 1.29 Consider the following sequences in R2 . Do they converge? If

so, what is their limit?

1 1+n 1 √
xn = ( 1+n , 3+2n ), yn = ( n1 , (−1)n ), zn = (1, cos(n π2 )), vn = n ( n, 1 + n1 ).

11
 We now use norms to define the continuity of a function f : Rd → R,
that is, a function of several variables. An example is the utility function of
a consumer, representing the preferences of a consumer in an economy with d
consumption goods. For instance, the Cobb-Douglas function with two goods
1−α
(d = 2): f (x1 , x2 ) = xα
1 x2 , 0 < α < 1. Thanks to the notion of sequence in
d
R that we just saw, we will obtain a useful characterization of continuity for
functions of several variables.

Definition (continuous function) A function f : Rd → R is continuous at

x ∈ Rd if

∀ε > 0, ∃δ > 0, ∀y ∈ Rd , k x − y k≤ δ ⇒| f (x) − f (y) |≤ ε. (6)

Function f is continuous on set A ⊆ Rd if it is continuous at all points x ∈ A.

Exercise 1.30 ? Let f : Rd → R be a linear function, defined by f (x) =

Pd
pi xi = p · x (wherep = (pi )1≤i≤d ).
i=1

1. Find a constant c > 0 such that for all x ∈ Rd , | f (x) |≤ c k x k. Hint:

choose cleverly the underlying norm.
2. Conclude that f is continuous.

Proposition (Sequential characterization of continuity) Function f : Rd → R

is continuous at x ∈ Rd if and only if for all sequences xn in Rd converging
towards x, f (xn ) converges towards f (x).

Proof
Necessary condition (⇒): assume f continuous at x, that is (??). Let xn
be a sequence converging towards x. We want to show that f (xn ) converges
towards f (x). That is, by (??), find, for all ε > 0, an integer n0 ∈ N such that,
for all n ≥ n0 , | f (xn ) − f (x) |≤ ε. Fix ε > 0. By (??), there exists δ = δ(ε)
so that for all y ∈ Rd , k x − y k≤ δ(ε) guarantees that | f (x) − f (y) |≤ ε.
Moreover, since xn converges towards x, by (??), for this δ(ε), there exists an
integer n0 = n0 (δ(ε)) such that, for all n ≥ n0 (δ(ε)), k xn − x k≤ δ(ε). Thus
n ≥ n0 (δ(ε)) guarantees | f (xn ) − f (x) |≤ ε.
Sufficient condition (⇐): We need to show that if, for all sequences xn
converging towards x, f (xn ) converges towards f (x), then (??) holds (an im-
plication, of type P ⇒ Q). We prove it by contraposition (that is, we prove Q
false ⇒ P false, which is equivalent to P ⇒ Q).
Assume thus that (??) does not hold. That is, there exists ε0 > 0 such that
for all δ > 0, there exists y ∈ Rd such that k x − y k≤ δ but | f (x) − f (y) |> ε0 .
1
Give ε0 , let δ = 1+n , for each n ∈ N: there exists yn ∈ Rd such that k x − yn k≤
1
1+n but | f (x) − f (yn ) |> ε0 . We found a sequence yn converging towards x,
but such that f (yn ) does not converge towards f (x).

12
Proposition (Properties of continuous functions) Let f : Rd → R and g : Rd →
R be functions that are continuous at x ∈ Rd , and a, b be real numbers. The
following functions are continuous at x: |f |, af + bg, f g, max {f, g}, min {f, g}
and, if g(x) 6= 0, fg . If g : R → R is continuous at f (x), then g ◦ f is continuous
at x.

Remark: we may need to consider mappings f : Rd → Rp , where p is an integer

> 1. The notions introduced may be applied to this case, as such a mapping f
is defined by p real-valued functions fj , j = 1, ..., p, with fj : Rd → R. For all
x ∈ Rd , f (x) ∈ Rp is of the form f (x) = (fj (x))1≤j≤p . Continuity of function
f : Rd → Rp at x ∈ Rd may be defined in two equivalent ways. A first definition
is that f is continuous at x if and only if each of the functions fj , j = 1, ..., p,
is continuous at x. The continuity of f at x may also be defined directly, as in
(??), using the distance in Rp between f (x) and f (y):

∀ε > 0, ∃δ > 0, ∀y ∈ Rd , k x − y k≤ δ ⇒k f (x) − f (y) k≤ ε.

Exercise 1.31 ? Check that the two above definitions are indeed equivalent.

13