4.

Single-variable optimisation

Julia Wirtz

ECON10005 - Mathematics for Economics

4. Single-variable optimisation ECON10005 1 / 26

Optimisation - Introduction

Economic approach to behaviour of economic agents:

Assumption 1: Objective of economic agents can be represented by a function.
Assumption 2: Agents make optimal choices over a set of feasible actions.

Example 1: Consumer behaviour

Objective of the consumer can be summarised by a utility function

Agent will maximise u(x1 , · · · , xn ) by choosing a specific consumption bundle

(x1 , · · · , xn ) from a set of feasible choices.

4. Single-variable optimisation ECON10005 2 / 26

Example 2: Firm behaviour
Objective of the firm can be summarised by a profit function

Firm will maximise π(y1 , · · · , yn ) by choosing prices/ quantities/ inputs

(y1 , · · · , yn ) from a set of feasible choices.

Example 3: Government behaviour

Objective of the government can be summarised by a welfare function

Government will maximise v(τ1 , · · · , τn ) by choosing tax rates (τ1 , · · · , τn ) from a

set of feasible choices.

4. Single-variable optimisation ECON10005 3 / 26

Formal Model

The baseline economic agent solves

max f (x1 , · · · , xn )
x1 ,··· ,xn

subject to possible constraints

f (x1 , · · · , xn ): target function

x1 , · · · , xn choices
constraints → determine feasible choices

4. Single-variable optimisation ECON10005 4 / 26

Preview

In the next lectures we will study

This lecture:
maxx∈S f (x): single variable optimisation.

Lecture 5:
max(x1 ,··· ,xn )∈S n f (x1 , · · · , xn ): multivariate optimisation.

Lecture 6:
max(x1 ,··· ,xn )∈S n f (x1 , · · · , xn ), subject to g(x1 , · · · , xn ) = 0:
multivariate optimisation with equality constraints.

Lecture 7:
max(x1 ,··· ,xn )∈S n f (x1 , · · · , xn ), subject to g(x1 , · · · , xn ) ≤ 0:
multivariate optimisation with inequality constraints.

4. Single-variable optimisation ECON10005 5 / 26

Single variable optimisation

Consider a function f defined over a domain D:

D −→ f (D)
x −→ f (x)

We want to find an ”action” x ∈ D which maximises (or minimises) the value f (x).

4. Single-variable optimisation ECON10005 6 / 26

Definition of a maximum/ minimum

f (x)

Maximum
A maximum point for f on D is a point c ∈ D such that
f (c) ≥ f (x), for all x ∈ D

x
f (x)

Minimum
A minimum point for f on D is a point c ∈ D such that
f (c) ≤ f (x), for all x ∈ D

x
4. Single-variable optimisation ECON10005 7 / 26
Finding a maximum/ minimum

Finding a maximum/ minimum is not easy.

We will see that, under some conditions, a maximum/ minimum is either
▶ a stationary point,
▶ a corner point,
▶ or non-existent.

4. Single-variable optimisation ECON10005 8 / 26

Stationary and corner point

Stationary point
f (x)
A stationary point of f is a point
c ∈ D such that:

f ′ (c) = 0

→ f needs to be differentiable

Corner point
If D = [a, b] is an interval, the
corner points are a and b. x
a b
→ requires finite bounds

4. Single-variable optimisation ECON10005 9 / 26

Global and local maximum/minimum
A point can be a local maximum but not a global maximum (i.e. THE
maximum).
A point can be a local minimum but not a global minimum (i.e. THE
minimum).

f (x)
MAXIMUM

local max

local min
local min

MINIMUM

x
a b

4. Single-variable optimisation ECON10005 10 / 26

Types of stationary points: Local minimum

Consider f twice differentiable over D

and c a stationary point [f ′ (c) = 0].
f (x)

c is a local minimum if
a) f ′′ (c) > 0
b) OR f ′ (c− ) < 0 and f ′ (c+ ) > 0 x
(f ′ is negative just below c and positive just
above)

4. Single-variable optimisation ECON10005 11 / 26

Types of stationary points: Local maximum

Consider f twice differentiable over D

and c a stationary point [f ′ (c) = 0].
f (x)

c is a local maximum if
a) f ′′ (c) < 0
b) OR f ′ (c− ) > 0 and f ′ (c+ ) < 0 x
(f ′ is positive just below c and negative just
above)

4. Single-variable optimisation ECON10005 12 / 26

Types of stationary points: Inflection point

Consider f twice differentiable over D

and c a stationary point [f ′ (c) = 0]. f (x)

c is an inflection point if
1. f ′′ (c) = 0
(
f ′′ (c− ) x
2. AND of different signs.
f ′′ (c+ )

4. Single-variable optimisation ECON10005 13 / 26

Summary: Types of stationary points

Consider f twice differentiable over D

and c a stationary point [f ′ (c) = 0]. f (x)
b
c is a:

a local minimum if f ′′ (c) > 0

b local maximum if f ′′ (c) < 0

c
c an
( inflection point if f ′′ (c)= 0 and a
′′
f (c ) −
of different signs.
f ′′ (c+ )
x

4. Single-variable optimisation ECON10005 14 / 26

Existence of a maximum/ minimum

When f is continuous over a closed, bounded interval [a, b],

a maximum and a minimum must exist.

Extreme value theorem

Consider f continuous over the interval [a, b]. There exist {c, d} such that:

f (c) ≤ f (x) ≤ f (d), for all x ∈ [a, b]

↑ ↑
minimum maximum

4. Single-variable optimisation ECON10005 15 / 26

Domain not bounded or not closed
When the domain of f is not bounded or or not closed,
there may be no maximum/ minimum.

Example: f (x) = ex

Closed & bounded interval: For x ∈ [a, b] the

f (x)
maximum/ minimum is at the corner points.

Not bounded: For x ∈ R neither the maximum

nor the minimum exist.

Not closed: For x ∈ (a, b) neither the

maximum nor the minimum exist.

f (x) increases as x gets bigger/ decreases as x

gets smaller, but there is not one value for x at x
a b
which f is maximised/ minimised.
4. Single-variable optimisation ECON10005 16 / 26
Function not continuous

When f is not continuous, there may be no maximum/ minimum.

f (x)

1
Example: f (x) = x with x ∈ [a, b]

f is not continuous at 0. a
x
Neither the maximum nor the minimum b
exist.

4. Single-variable optimisation ECON10005 17 / 26

Necessary condition for a maximum/ minimum

How to find a maximum/ minimum?

When a maximum/ minimum exists,

→ then it must be either a stationary or a corner point.

Necessary condition
Consider f differentiable over the interval [a, b].
If c is a maximum (or minimum) of f on [a, b], then:

a) either f ′ (c) = 0 [stationary point]

b) or c = a, or c = b [corner point]

4. Single-variable optimisation ECON10005 18 / 26

Compare all stationary and corner points

f (x)
MAXIMUM

local max

local min
local min

MINIMUM

x
a b

4. Single-variable optimisation ECON10005 19 / 26

Finding a maximum/ minimum: f differentiable over [a, b]
Method for closed and bounded domain [a, b]

1. Find all stationary points: f ′ (c) = 0 over [a, b]

2. Evaluate the function f at all candidates

a) the stationary points
b) the corner points a and b.

3. Maximum: Largest value among the candidates

Minimum: Smallest value among the candidates

Remarks:
Corner solution: If maximum/ minimum is a corner point (a or b)
Interior solution: otherwise
There may be several maxima/ minima

4. Single-variable optimisation ECON10005 20 / 26

Concave functions

f (x)

If function f
1. is concave and
2. has a stationary point x
f (x)
Then: The stationary point is the
maximum

x
4. Single-variable optimisation ECON10005 21 / 26
Convex functions

f (x)

If function f
1. is convex and
2. has a stationary point x
f (x)
Then: The stationary point is the
minimum

x
4. Single-variable optimisation ECON10005 22 / 26
Summary: Convex/ concave functions

Proposition: Convex/ concave

Consider f twice differentiable over [a, b], and c a stationary point.

If f is concave, c is a maximum.
If f is convex, c is a minimum.

4. Single-variable optimisation ECON10005 23 / 26

Finding a maximum/ minimum: Domain not closed or bounded
E.g. f is differentiable on (−∞, ∞)

1. Necessary condition: Find all stationary points: f ′ (c) = 0

2. Possible sufficient conditions:

Maximum Minimum
a) f concave: f ′′ < 0 a) f convex: f ′′ > 0

b) f ′ > 0 for all x < c and b) f ′ < 0 for all x < c and
f ′ < 0 for all x > c f ′ > 0 for all x > c

c) f (c) ≥ limx→±∞ c) f (c) ≤ limx→±∞

as well as largest among stationary as well as smallest among stationary
points points

4. Single-variable optimisation ECON10005 24 / 26

Economic application: A monopoly

Monopoly produces goods at cost per unit c > 0.

Demand: D(p) = e−kp , with k > 0.

Objective: Set price to maximise profit.

4. Single-variable optimisation ECON10005 25 / 26

Reading

Sydsaeter et al. (2016). Essential mathematics for economic analysis, ch. 8.

4. Single-variable optimisation ECON10005 26 / 26