Lecture 1 Introduction To Optimization in Economics
Lecture 1 Introduction To Optimization in Economics
(4SSMN903)
Lecture 1
Eddy Tam
January 9, 2022
Optimization
Definition
(Local maximum) A point x ∗ is local maximum if for some > 0,
x ∗ − < x < x ∗ + (with x in the domain), f (x ∗ ) ≥ f (x )
I It is a strict global maximum if strict inequality holds (i.e.
f (x ∗ ) > f (x ) for any x in the domain); similarly for local
maximum.
Global and Local Maximum
Definition
(Local minimum) A point x ∗ is local minimum if for some > 0,
x ∗ − < x < x ∗ + (with x in the domain), f (x ∗ ) ≤ f (x )
Existence of maximizer
Theorem
(Existence of maximizer)
Continuous function in a closed and bounded domain has a
maximum (and minimum).
I Closed domain (single variable) - the set include the end
points of interval. (e.g. −2 ≤ x ≤ 2, as opposed to
−2 < x < 2)
I Bounded domain (single variable) - there is an upper bound
and lower bound. (e.g. −2 ≤ x < ∞ is not a bounded
domain)
I e.g. There is a maximum for y = x 2 for −2 ≤ x ≤ 2. Is there
a maximum for y = x 2 if x takes any value from
−∞ < x < ∞?
Existence of maximizer
e.g. Is there a maximum for the function y = x 2 for −2 ≤ x ≤ 2?
What about if x takes any value from −∞ < x < ∞?
Theorem
(Necessary condition for local extreme point) If an interior point x ∗
is a local maximum or minimum point for a differentiable function
f (x ), then its first order derivative f 0 (x ) is zero at x ∗ , i.e.
f 0 (x ∗ ) = 0
We set f 0 (x ∗ ) = −2x ∗ = 0 =⇒ x ∗ = 0.
The first order condition suggest that if there is an interior point
that is a local maximum or minimum, it would be x ∗ = 0.
First order condition
For y = f (x ) = x 2 , f 0 (x ) = 2x
We set f 0 (x ) = 2x = 0 =⇒ x = 0.
The first order condition suggest that if there is an interior point
that is a local maximum or minimum, it would be x ∗ = 0.
Example
√
I We assume a production function q = F (L) = L with labor
input L, price of output is p, quantity produced is q, wage
rate is w .
√
I maxL π(L) = p L − wL for 0 ≤ L < ∞
1. First order condition is
∂π 1 1
= p L− 2 − w = 0
∂L 2
p 2
=⇒ L∗ = ( )
2w
p 2
2. It implies L∗ = ( 2w ) that satisfy the first order condition is
candidate of local maximum or local minimum.
Is the first order condition sufficient?
Definition
For a function f (x ), a point x ∗ at which the derivative is equal to
zero, i.e. f 0 (x ∗ ) = 0, is called stationary point.
I Stationary point may be an extreme point or point of
inflection
Stationary point
Theorem
For function f (x ) that is twice differentiable,
I If f 0 (x ∗ ) = 0 and f 00 (x ∗ ) < 0, x ∗ is a local maximum point
I If f 0 (x ∗ ) = 0 and f 00 (x ∗ ) > 0, x ∗ is a local minimum point
Definition
Let x and y be two points on the domain. A function f (x ) is
concave if for all x , y in the domain, with t in [0,1].
It is strictly concave if
Definition
Let x and y be two points on the domain. A function f (x ) is
convex if for all x , y in the domain, with t in [0,1].
It is strictly convex if
Theorem
If f (x ) is a concave function defined on an interval, and if x ∗ is an
interior point and f 0 (x ∗ ) = 0, x ∗ is a maximum point for f
Theorem
If f (x ) be a convex function defined on an interval, and if x ∗ is an
interior point and f 0 (x ∗ ) = 0, x ∗ is a minimum point for f
I If the objective function is concave, the interior point that
satisfies the first order conditions is a global (and local)
maximum point.
Concave and convex functions in economics
3. AC 00 (q) = 2F
q3
> 0 for q > 0
4. As AC 00 (q)
> 0 for all q in the domain, the average
q cost
∗ F
function is strictly convex. This implies q = c2 is a
minimum for the average cost function.
Implication of concave function in economics - example
1. Suppose a consumer with utility function U(x ) = ln(x ) face a
lottery that gives 2 unit of goods with probability 12 , and 8
unit of goods with probability 21 . The consumer need to
choose between the lottery, or getting 5 unit of goods for
certain.
I The lottery gives expected amount of 21 2 + 12 8 = 5
I If assume further that consumer make decision based on
expected utility, i.e. the lottery gives the consumer utility
1 1
2 ln(2) + 2 ln(5)
I Since we know the function ln(x ) is concave, this means
1 1
ln(5) ≥ ln(2) + ln(5)
2 2
I The consumer prefer having 5 for certain than choosing the
lottery.
I As ln(x ) is concave, the consumer prefer having for certain the
expected value of any lottery than the lottery itself. We
consider the consumer as risk-averse.
Implication of convex function in economics - example
1. Consider a firm face a cost function C (q) = q 2
I The first derivative is ∂C
∂q = 2q.
I It is the marginal cost of production - the cost for producing
an additional unit at the margin.
I As production increase, how does the marginal cost change?
I
∂MC ∂2C
= =2>0
∂q ∂q 2
I The marginal cost is increasing with quantity produced - the
more is the production, the more expensive to produce the
additional amount.
I It could be because one/some of the input is fixed, it become
more expensive to scale up production (e.g. with a manager, it
need 2 workers to produce a good, when a manager is
occupied, it need 4 workers to produce an additional good).
I As C 00 (q) > 0, the (strict) convexity in cost function also
implies increasing marginal cost.
Maximization under a specific domain - Applications in
economics
Theorem
Let the differentiable function f (x ) be defined over the interval
a ≤ x ≤ b. If x ∗ is a minimum for f (x ) over a ≤ x ≤ b, at least
one of the following conditions would hold
1. f 0 (x ∗ ) ≥ 0 and (x ∗ − a)f 0 (x ∗ ) = 0
2. f 0 (x ∗ ) ≤ 0 and (b − x ∗ )f 0 (x ∗ ) = 0
s.t. 0.1 ≤ q ≤ 2
2. Calculate the first order derivative: f 0 (q) = 6q 2 − q
3. Search for an interior solution by finding the stationary
point(s)
I Set first order derivative to zero:
6q 2 − q = 0 =⇒ q(6q − 1) = 0
I q = 0 or q = 61 (rule out q = 0 as it is not in the domain)
Profit maximization
for 0 ≤ x ≤ mp
3. First order condition is
∂ Ũ 1
= −p =0
∂x x +1
1
=⇒ =p
x +1
4. It suggest the marginal utility ∂U(x
∂x
,y )
= 1
x +1 from x equals the
price the consumer has to pay for x
Application 4 - Utility maximisation
that x ∗ = 1
p − 1 is a local maximum.
7. x∗ 1
= − 1 is in the domain if 0 ≤
p
1
p −1≤ m
p, or when p ≤ 1
and 1 ≤ m + p
8. The derivative of the objective function Ũ 0 (x ) at the two end
points are Ũ 0 (0) = 1 − p and Ũ 0 ( m 1
p ) = p( m+p − 1)
9. If p > 1, Ũ 0 (0) < 0, x = 0 is candidate of maximum. (x is
"too expensive")
10. If m + p < 1, Ũ 0 ( m
p ) > 0, x =
m
p is a candidate of maximum.