ECC2000 S2/2018

Intermediate microeconomics

Tutorial 3

1 Optimization
• “Optimization” is the “quest for the best” - i.e. the quest for the min/max=extremum.

– We start with an objective function

∗ the dependent variables represent the object of maximization or minimization
∗ the independent variables represent the choice variables
∗ the economic unit in question: finds the set of values for the choice variables
that would lead us to the desired extremum of the objective function

Let y = f (x) and assume it is continuously differentiable - We will start by developing

procedures for finding the level of x that maximizes or minimizes the value of y.

1
1.1 Relative MAX and MIN: First-Derivative Test
• Relative (local) extremum: A point x0 is a relative extremum if it represents the
highest (max)/lowest (min) point of all the points in the neighbourhood of x0 .

REMARK: The absolute maximum is the HIGHEST relative maximum or one of the end
points of a function. The absolute minimum is the LOWEST relative minimum or one of
the end points of a function.

2
The first derivative f 0 (x) will play a major role in finding the extremum values. - Why? See
graph.

1st derivative test: Let y = f (x) ∈ C (1) , if x = x0 is a relative extremum then f 0 (x0 ) = 0
(first-order condition: FOC) and
1) x0 is a relative maximum if f 0 (x0 ) changes sign from + to − from the immediate
left of x0 to its immediate right, or
2) x0 is a relative minimum if f 0 (x0 ) changes sign from − to + from the immediate
left of x0 to its immediate right,
3) x0 is an inflection point if f 0 (x0 ) maintains a constant sign.

Then x0 is the Critical value/point, f (x0 ) is the Stationary value of y, and

(x0 , f (x0 )) is the Stationary Point.

• f 0 (x0 ) = 0 is the necessary condition, and the change of sign is the sufficient condition
for existence of a relative extremum x0 (relative max or min).

• For this course, we focus on the necessary condition, or the first order condition (FOC).

3
2 Optimization with equality constraints
2.1 Effects of constraints
Let U (x1 , x2 ) = x1 x2 + 2x1 be the utility function over two goods with M Ui = ∂U/∂xi > 0,
∀xi > 0, i ∈ {1, 2}
To maximize U , a consumer will need to purchase ∞ amounts of x1 and x2 , BUT there is a
problem of RESOURCES → need to take into consideration the agents purchasing power
→ budget constraint: given the agents income (how much willing to spend on consumption)
and given prices P1 and P2 , how much of goods x1 and x2 can the agent consume (2-good
example). For example:

4x1 + 2x2 = $60, where P1 = 4 and P2 = 2

The problem then becomes

maxU = x1 x2 + 2x1
x1 ,x2

s.t. 4x1 + 2x2 = 60

which will narrow the domain (the values that x1 and x2 can take) and narrow the range of
the objective function U . Furthermore, x1 and x2 will be dependent on each other.
Solving the problem

4x1 + 2x2 = 60 =⇒ x2 = 30 − 2x1 substitute into U

=⇒ max U = x1 (30 − 2x1 ) + 2x1 = 32x1 − 2x21
x1

4
2.2 Method of Lagrange Multiplier
Often we can encounter multiconstraints or constraints which are complicated functions,
making it difficult to use the technique of substitution and elimination of variables. There-
fore, we can use the method of Lagrange to solve a constrained optimisation problem:

maxU = x1 x2 + 2x1
x1 ,x2

s.t. 4x1 + 2x2 = 60

Writing up the Lagrangian function:

L = x1 x2 + 2x1 + λ [60 − 4x1 − 2x2 ]

where λ denotes the Lagrange multiplier. By introducing the Lagrange multiplier as an
additional variable — we can apply to the constrained-extremum problem the same first-
order conditions used in the free-extremum problem.

Finding the FOC:

Generally, the constrained optimisation problem using the method of Lagrange:

maxz = f (x, y)
x,y

s.t. g(x, y) = c

L = f (x, y) + λ [c − g(x, y)]

FOC:  ∂L

 ∂λ
=c− g(x, y) = 0 )
∂L
∂x
= fx − λgx = 0 fx fy
 ∂L
=⇒ gx
= gy
=λ
= fy − λgy = 0

∂y

5
3 Uncertainty
An investor is considering whether to purchase Microsoft stock. The investor’s utility func-
1
tion is U (r) = r 2 , where r is the return that the investor earns. The return is uncertain,
and the table below shows the distribution of returns.
r Probability
0.01 0.2
0.04 0.4
0.09 0.3
0.25 0.1

1. Calculate the expected return.

2. Calculate the expected utility.

3. Calculate the certainty equivalent.

4. Calculate the risk premium.

5. What is this investor’s attitude towards risk?

6
Appendix: THE LAGRANGE MULTIPLIER
INTERPRETATION OF THE LAGRANGE MULTIPLIER: The Lagrange multi-
plier, λ∗ , measures the sensitivity of L∗ to changes in the constraint:
Let c be exogenous, since λ, x, y are endogenous → ∆c will imply a shift in the constraint
curve that will lead to a different solution. Hence, an ↑ c → relaxation of the constraint!

Using the Implicit Function Theorem on the 3 equations

∂L
F1 (λ, x, y; c) = = c − g(x, y) = 0,
∂λ
∂L
F2 (λ, x, y; c) = = fx − λgx = 0,
∂x
∂L
F3 (λ, x, y; c) = = fy − λgy = 0
∂y
We assume there exists a solution to the above system of three equations, λ∗ = λ∗ (c),
x∗ = x∗ (c), y ∗ = y ∗ (c), with continuous partial derivatives such that

∗ ∗
c − g (x , y ) = 0

fx (x∗ , y ∗ ) − λ∗ gx (x∗ , y ∗ ) = 0

fy (x∗ , y ∗ ) − λ∗ gy (x∗ , y ∗ ) = 0


How does the optimal value change with c?

L∗ = L∗ (c) = f (x∗ , y ∗ ) + λ∗ [c − g (x∗ , y ∗ )]

Take full derivative w.r.t. c,

∂L∗ dx∗ dy ∗ ∗
dx∗ dy ∗
 
∗ ∗ dλ ∗
= fx + fy + [c − g (x , y )] + λ 1 − gx − gy
∂c dc dc dc dc dc
∗ ∗ ∗
dx dy dλ
= (fx − λ∗ gx ) + (fy − λ∗ gy ) + [c − g (x∗ , y ∗ )] + λ∗
dc dc dc
Apply FOCs
∂L∗
=⇒ = λ∗
∂c
OR
∂f (x∗ , y ∗ ) dx∗ dy ∗
= fx (x∗ , y ∗ ) + fy (x∗ , y ∗ ) [Apply F OCs]
∂c dc dc
∗ ∗
∗ ∗ ∗ dx ∗ ∗ ∗ dy
= λ gx (x , y ) + λ gy (x , y )
dc dc 
∗ ∗

dx dy
= λ∗ gx (x∗ , y ∗ ) + gy (x∗ , y ∗ )
dc dc
dx∗ dy ∗
Take full derivatives on:c − g (x∗ , y ∗ ) = 0 =⇒ gx (x∗ , y ∗ ) + gy (x∗ , y ∗ ) =1
dc dc
∂f (x∗ , y ∗ )
subsitute into above =⇒ = λ∗
∂c

7
=⇒Therefore, we have demonstrated that the Lagrange Multiplier, λ∗ , indeed measures the
effect of a change in the constraint via “c” on the optimal value of the objective function.
Often λ∗ is referred to as the shadow price.
In this course, it means that the Lagrange Multiplier, λ∗ , measures the effect of a change in
income on the optimal value of the objective function.