Single Variable Optimization

[SHS] 8.1 – 8.7

1
Single variable optimization – An example
No Arbitrage Condition
 Optimization with constraints
 Single choice variable

 Profit maximizing

Max π (Q)
Q

s.t. 0 ≤ Q ≤ Q
where Q is the capacity of the firm.

 Optimal solution: Q *
 Impossibility of finding an improvement -- No Arbitrage

π (Q*) ≥ π (Q) for any Q ∈ [0, Q ]

 Optimal solution <==> No Arbitrage

2
Marginal analysis -- No Arbitrage Locally -- FOC
 Infinitesimal change -- dQ -- marginal analysis -- Calculus
 Optimal solution ==> No Arbitrage locally

 Interior points: zero gradient (slope) -- stationary points

 No Arbitrage locally ==> stationary points

dπ
π '(=
Q) = 0
dQ
 first order (necessary) condition – FOC
 Plus points where marginal analysis NA – corner points: 0 and Q

 Optimal solution Q* is in the set: {stationary points, corner points}

 Interior solution vs. Corner solution

 Interior solution: FOC -- π '(Q*) = 0

 Corner solution: FOC -- NA

3
 Second order (sufficient) condition – SOC
 Local optimum – neighborhood
 Local maximum vs. local minimum

 Characteristic of π (Q ) -- “locally” strictly concave or convex

 second order (sufficient) condition – SOC

d 2π
 π ''(=
Q) Q =Q* < 0 ==> Q * Local maximum
dQ 2 Q =Q*

d 2π
 π ''(=
Q) Q =Q* >0 ==> Q * Local minimum
dQ 2 Q =Q*

 Global optimum
d 2π
 Single peaked -- concave everywhere: π ''(=
Q) 2
≤ 0 for all Q
dQ
 stationary points from FOC ==> Global maximum

4
The Optimization Problem
 The optimization problems take the form:
Max f ( x)
x

s.t. x ∈ S
• f is the objective function (utility, profit, …)
• x is the choice variable (consumption bundle, input combination, …)
• S is the constraint set (the set of feasible actions)

 Suppose that the value x* solves the problem above. Then f (x*) ≥ f (x)
for all x ∈ S. We say that x* is a (global) maximizer (or maximum
point) of the function f subject to the constraint.
 f (x*) is the (global) maximum (or maximum value) of the function f
subject to the constraint.
1. Existence?
2. If it exists, how to identify it?

5
 Extreme Value Theorem
 Extreme value theorem
A continuous function on a compact interval attains both a
maximum and a minimum on the interval.
 compact -- closed and bounded

 Example: f (x) = x, S = [0, ∞).

f increases without bound, and never attains a maximum.
 Example: f (x) = 1 − 1/x, S = [1, ∞).
f converges to the value 1, but never attains this value.
 Example: f (x) = x, S = (0, 1).
The points 0 and 1 are excluded from S. As x approaches 1, f (x)
approaches 1, but this value is never attained for values of x in S.
 Example: f (x) = x if x < 1/2 and f (x) = x − 1 if x ≥ 1/2; S = [0, 1].
As x approaches 1/2 (from left), f (x) approaches 1/2, but this
value is never attained, because at x = 1/2 the function jumps
down to −1/2.
6
Global Maximum vs. Local Maximum
 Example:

− x2
Max f ( x=
) ( x − 0.5) e 2
x

s.t. x ∈ [−2, 2]

 The point x* ∈ S is a local maximizer (or maximum point) if f(x) ≤ f(x*)

for the neighborhood of x*. That is, there is an open interval
( x * −ε , x * +ε ), ε > 0
such that f ( x) ≤ f ( x*) for all x ∈ ( x * −ε , x * +ε ) .
The value f(x*) is called the local maximum (or maximum value) of f.
 ε-neighbourhood of x* -- open interval centered at x* with radius ε.

7
 Minimization problems
 A global minimizer (or minimum point) and a local minimizer (or
minimum point) are defined analogously.
 minimum (or minimum value) of f

 local minimum (or minimum value) of f

 Any minimization problem can be turned into a maximization problem

by taking the negative of the objective function.

 That is Min f ( x)
x

s.t. x ∈ S
is equivalent to
Max − f ( x)
x

s.t. x ∈ S
 Then apply the results for maximization problems.
8
 First-order condition
 Example:
Max f ( x) =2 + ( x − 3)3
x

s.t. x ∈ [1, 5]
 FOC: necessary condition
 Inflection point

 A stationary point is not necessarily a global maximizer, or even a local

maximizer, or even a local optimizer of any sort (maximizer or minimizer).

 Example Max f ( x) = 1 − x
x

s.t. x ∈ [0,1]
 FOC: no stationary point!
 A global maximizer is not necessarily a stationary point.
 However, if interior points are max or min points, then the maximum

or minimum point is a stationary point.

9
First-order condition
 The first-order (“necessary”) condition:
Let f be a differentiable function defined on the interval I=[a,b]. If an
interior point x* ∈ I is a local or global maximizer or minimizer of f ,
then f '(x*) = 0.

 Interior Solution: optimal solution is an interior point of I

 f '(x*)=0

 Corner Solution: optimal solution is NOT an interior point of I

 FOC -- NA

10
 First-order Condition -- example
 Example
max x f ( x) =( x − 1)( x − 2)( x + 3)
s.t. x ∈ [−5,5]
FOC f '( x)= 3 x 2 − 7= 0
7 7
x1 = x2 = −
3 3
14 7
f ( x1 )= 6 −
3 3
14 7
f ( x2 )= 6 + f (−5) =−84 f (5) = 96
3 3
 Corner solution: x*=5, where f '( x) x =5− > 0
− x2
 Exercise Max f ( x=
) ( x − 0.5) e 2
x

s.t. x ∈ [−2, 2]
11
Second-order condition -- local optimum
 Suppose we find the stationary points from FOC.
 How can we tell whether a stationary point is a local maximum, a

local minimum, or neither?

 examining the second derivative at the stationary point

 Recall that for twice-differentiable functions over an interval I

f ''( x) < 0 ⇒ f strictly concave
f ''( x) > 0 ⇒ f strictly convex

 The second-order “sufficient” condition

Let f be a twice differentiable function defined on the interval I.
Suppose an interior point x* ∈ I is a stationary point of f , (so that
f '(x*) = 0). Then
(i) If f "(x*) < 0 then x* is a local maximizer.
(ii) If f "(x*) > 0 then x* is a local minimizer.
 “locally” strictly concave or “locally” strictly convex: f "(x*)

12
 Second-order condition -- local optimum
 Example maxx f (x) = x2 + 3 s.t. x ∈ [−1, 1]

 Exercise maxx f (x) = −x3 + 3x − 2 s.t. x ∈ [−2, 2]

 Example maxx f (x) = x3 f ' (0) = f "(0) = 0

 Example maxx g (x) = x4 g ' (0) = g"(0) = 0

 Example maxx h(x) = − x4 h' (0) = h "(0) = 0

 If f "(x*) = 0 then the condition above cannot tell, without further

investigation, whether x* is a local maximizer or local minimizer of f ,
or neither.

13
 Sufficient conditions for global optimum
 Single peaked – concave everywhere
 stationary points from FOC ==> Global maximum
 Recall that for twice-differentiable functions over an interval I
f ''( x) ≤ 0 ⇔ f concave
f ''( x) ≥ 0 ⇔ f convex
 Single peaked – concave everywhere: f "(x) ≤ 0 for all x

 Sufficient conditions for global optimum

Let f be a differentiable function defined on the interval I. Then
(i) if f is concave, i.e. f "(x) ≤ 0 for all x,
x* is a (interior) stationary point of f ==> x* is a global maximizer of f in I
(ii) if f is convex, i.e. f "(x) ≥ 0 for all x,
x* is a (interior) stationary point of f ==> x* is a global minimizer of f in I.

 Example maxx −x2 s.t. x ∈ [−1,1]

14
Procedure for finding the maximum -- a differential
function f defined on an interval I
Step 1. Find all the stationary points of f (points x for which f '(x) = 0)
Step 2. Check the Second-order condition -- sufficient condition for
global optimum. If it is satisfied, we are done!
 If not, continue to the following steps.

(May check the SOC for the local optimum.)

Step 3. Extreme Value Theorem (EVT), check if the object function f is

continuous and the interval I is compact.
 If they are, the only points that can be global maximizers are either
stationary points or boundary points of the interval I.
Step 4. Find the values of f at the endpoints, if any, of I.
Step 5. Compare, the points you have found at which the value of the
objection function is the largest are the maximizer.

 Interior Solution vs. Corner Solution

15
Application -- linear calling plan
 Joel currently participates into a calling plan from SingTel. The
policy of the plan is: $0.1 per each minute of outgoing talks. There is
no monthly subscription charge. His monthly benefit function is

Benefit = 10.1Q − 0.05Q 2

where Q is the minutes of outgoing talks. The measurement unit of
benefit is $.

Find out the optimal amount of minutes of outgoing talks.

16
 Application -- Profit Maximization (Monopoly –
linear demand and quadratic cost)
Suppose a monopolist sets its price p, when the (inverse) market demand is
p = 1200 − 2q
where p is market price and q is quantity demand. Its cost function is
C(q) = 100+ q2
where q is the output of the monopolist. In addition, the capacity of the firm
is 1000, which means q≤1000.
(a) Find out the optimal output level, the market price, and the profit of the
monopolist.
 Profit-maximizing markup (price-cost margin): Lerner Index,

measure of market power

 difference between (profit maximizing) price and marginal cost,
expressed as a fraction of (profit maximizing) price:
markup = [p – MC] / p
(b) Find out the price markup for the monopolist at the optimal output level.

17
 Monopoly ==> Perfect competition
 (c) Suppose the entry barrier for the industry is removed. There are
more firms entering this market such that the industry becomes a perfect
competitive one. Assume all firms (including the previous monopolist)
have the same cost structure, i.e. C ( q=
i) 100 + qi 2 . Find out the number
of firms in this perfect competitive industry and the market price.

Hint:
 In perfect competitive industry, all firms are price taker.

 p=MR=MC

 Perfect competitive firms get zero profit in the long-run.

18
Comparative Statics &
Envelope Theorem

[SHS] 13.7

19
Example -- Profit Maximization (Price Taker)
 If a firm is a price taker, i.e., it has to take price as given, its profit
function is then π (Q) = R (Q) − C (Q) =PQ − C (Q)
 Assume C ′(Q ) > 0 and C ′′(Q ) > 0 .

Max π (Q) =R (Q) − C (Q) =PQ − C (Q)

Q

s.t. Q ∈ [0, Q]
 FOC π ′(Q ) = R ′(Q) − C ′(Q) =
P − C ′(Q) =
0
 Marginal Revenue = Marginal Cost
 Suppose interior solution exists ( Q is sufficiently large): Q* ∈ (0, Q) .
 Q* is implicitly defined by the FOC: P − C ′(Q*) = 0.
 SOC π ′′(Q ) = −C ′′(Q) < 0
 Maximum amount of profit: π (= Q*) PQ * −C (Q*)
 If price changes, …

P − C ′(Q*) =0 ⇒ Q* = Q * ( P)
π (Q * ( P)) =PQ * ( P) − C (Q * ( P)) =π * ( P)

20
 Comparative Statics
 Comparative statics is comparison of two different economic outcomes,
before and after a change in some underlying exogenous parameter.
1. How does Q*(P) change as P changes?
2. How does π*(P) change as P changes?
 Q* is implicitly defined by the FOC: P − C ′(Q*) = 0
 Implicit differentiation, take the derivative of FOC with respect to P.
dQ * dQ * 1
1 − C ′′(Q * ) =
0 ⇒ = > 0
dP dP C ′′(Q*)
 Law of supply

 Taking the derivative of the maximum amount of profit π (=

Q*) PQ * −C (Q*)
with respect to P,
d π (Q*) dQ * dQ * dQ*
= Q * +P − C ′(Q*) = Q * + [ P − C ′(Q*) ] = Q* > 0
dP dP dP dP
 Envelope theorem d π * ( P)
= Q* > 0
dP
 Numerical analysis ---> Symbolic analysis
21
 Comparative Statics & Envelop Theorem
 Consider a standard optimization problem where we wish to maximize
f(x; r), which is a function of the choice variable x, and a parameter r.
Max f ( x, r )
x
 The optimal choice x* is a function of the parameter r.
 The objective function at its optimum is called the value function:
=f ∗ (r ) max
= x f ( x , r ) f ( x ∗
( r ), r )

 When parameter r changes, there are two types of comparative statics:

1. the effect on the optimal choice x*(r)
2. the effect on the objective function f(x*(r),r)

 Envelope theorem df ∗ (r ) ∂f ( x, r )
= x = x∗ ( r )
dr ∂r
 The change in the maximal value of the function as a parameter changes
is simply the change caused by the direct impact of the parameter on the
function, holding the value of x fixed at its optimal value.
22
Envelop Theorem
 Envelop Theorem
df ∗ (r ) ∂f ( x, r )
= x = x∗ ( r )
dr ∂r
Proof
df ∗ (r ) ∂f ( x∗ (r ), r ) dx∗ (r ) ∂f ( x∗ (r ), r )
+
dr  ∂x dr  ∂r
The indirect effect on f through x ∗ The direct effects on f

∂f ( x∗ (r ), r ) ∂f ( x, r )
=
0+ = x = x∗ ( r )
∂r ∂r
 … the change in the maximal value of the objective function f(x,r) as
parameter r changes is simply the change caused by the direct
impact of the parameter r on the objective function f(x,r), holding the
value of x fixed at its optimal value x*(r) …
 the indirect effect, resulting from the change in the optimal value

of x*(r) is zero.
23
 Comparative Statics -- Optimal input choice
Example: A firm is a price taker, who pays w=2 for each unit of labor input.
It obtains the revenue p for each unit of output that it sells. Its output from l
units of the input is l .
(a) For what value of l is its profit maximized?
(Assume the upper bound of labor input usage is sufficiently large, such
that we have the interior solution.)
(b) If the parameter p changes, how does the optimal choice and maximum
value of the profit function change accordingly (in the marginal sense)?

Exercise:
max l π (l , w) = l − wl s.t. l ∈ [0, l ]
where w is a positive constant.
(a) Find out the optimal choice. (Assume the upper bound of labor input
usage is sufficiently large, such that we have the interior solution.)
(b) What is the effect of w on the optimal choice and maximum value of the
objective function (in the marginal sense)?

24