Mathematical knowledge

1 Functions of a single variable

• Function y = f (x)
• The first derivative of f with respect to x is

df
f 0 (x) = .
dx

It gives, at each value x, the slope or instantaneous rate of change in f (x).

• The second derivative of f with respect to x is

d2 f
f 00 (x) = .
dx2

It gives the rate at which the slope of f changes. It is thus related to the
curvature of the function f .
• Rules of differentiation

— For constants, α:

d
α = 0,
dx

— For sums:

d
[f (x) ± g(x)] = f 0 (x) ± g 0 (x),
dx

— Power rule:

d
(αxn ) = nαxn−1 ,
dx

1
 — Product rule:

d
[f (x)g(x)] = f 0 (x)g(x) + f (x)g 0 (x),
dx

— Quotient rule:

d f (x) f 0 (x)g(x) − f (x)g 0 (x)

[ ] = ,
dx g(x) [g(x)]2

— Chain rule:

d
[f (g(x))] = f 0 (g(x))g 0 (x)
dx

• Examples. Calculate the derivatives in each of the following cases:

1. y = 5x−2
2. f (x) = 2x2 + 3x + 4
3. g(x) = 3x − 1
4. (f (x)g(x))0 ?
5. ( fg(x)
(x) 0
)?
6. f (g(x))0 ?

• Concavity and first and second derivatives.

If f is twice differentiable, the following statements 1 to 3 are equivalent:

1. f is concave.

2. f 00 (x) ≤ 0 ∀x.

3. If λ ∈ [0, 1], ∀x and ∀x0 ,

f (λx + (1 − λ)x0 ) ≥ λf (x) + (1 − λ)f (x0 )

(For all x0 : f (x) ≤ f (x0 ) + f 0 (x0 )(x − x0 ) ∀x.)

• Moreover,.

2
 4. If f 00 (x) < 0 ∀x, then f is strictly concave.

• Examples. Are the following functions concave?

1. f (x) = 3x2 − 2x + 5
2. g(x) = −5x3 + 9x + 3
3. h(x) = 3x − 3

2 Functions of several variables (n=2)

• Function y = f (x1 , x2 )
• Partial derivatives

∂f
for i = 1, 2
∂xi

• Total differential dy of the function y = f (x1 , x2 )

∂f ∂f
dy = dx1 + dx2
∂x1 ∂x2

where dx1 and dx2 are small changes in x1 and x2 .

• Example: f (x1 , x2 ) = x21 + 3x1 x2 − x22 .

— What are the partial derivatives?

3 Optimization
• Function y = f (x) is differentiable.
• The function achieves a local maximum (respectively global maxi-
mum) at x∗ , if f (x∗ ) ≥ f (x) for all x in some neighborhood of x∗ (re-
spectively for all x).
• The function achieves a local minimum (respectively global minimum)
at xe, if f (e e (respectively for
x) ≥ f (x) for all x in some neighborhood of x
all x).

3
 • Necessary conditions for local interior optima in a single-variable case
f (x) is twice continuously differentiable. It reaches a local interior

1. maximum at x∗ ⇒ f 0 (x∗ ) = 0 (F OC)

⇒ f 00 (x∗ ) ≤ 0 (SOC)

e ⇒ f 0 (e
1. maximum at x x) = 0 (F OC)
00
⇒ f (e x) ≥ 0 (SOC)

4 Constrained optimization
• Consider the following problem

(
M axf (x1 , x2 )
x1 ,x2
subject to g(x1 , x2 ) = 0

• f (x1 , x2 ): objective function or maximand,

• x1 and x2 are choice variables,
• g(x1 , x2 ): constraint.

• Let solve this problem by substitution:

— suppose we can rewrite the constraint g(x1 , x2 ) = 0 as x2 = ge(x1 ).

— We can substitute this directly into the objective function: the 2
variable constrained maximization problem can be rephrased as the
single variable problem with no constraints:

M axf (x1 , ge(x1 ))

x1

— x∗1 is defined by

∂f (x1 , ge(x1 )) ∂f (x1 , ge(x1 )) de

g (x1 ))
+ = 0
∂x1 ∂x2 dx1

4
 — and x∗2 by

x∗2 = ge(x∗1 )

• Example: Cost minimizing firm (C = wL + rK) with Cobb-Douglas pro-

duction functions

M inwL + rK
K,L
subject to Q = Lα K β

1 β
— The constraint Q = Lα K β becomes L = Q α K − α .
— Unconstrained minimization

1 β
M inwQ α K − α + rK
K

— FOC:

dC 1 β − β −1
= −wQ α K α +r =0
dK α

— SOC

d2 C β 1 β β
= ( + 1)wQ α K − α −2 ≥ 0
dK 2 α α

— Solve FOC gives

w 1β α
K∗ = [ Q α ] α+β
r α

1 β
— Substitute K ∗ in L = Q α K − α gives

r 1α β
L∗ = [ Q β ] α+β
w β