5.
Multivariate optimisation
Julia Wirtz
ECON10005 - Mathematics for Economics
5. Multivariate optimisation ECON10005 1 / 13
Multivariate optimisation
Multivariate optimisation
Single-variable optimisation
f (x)
5. Multivariate optimisation ECON10005 2 / 13
Multivariate optimisation
Maximisation problem
max f (x1 , · · · , xn )
(x1 ,··· ,xn )∈S
Choice variables: (x1 , · · · , xn ) ∈ S
Target function: f : S −→ R
Similar to single-variable optimisation except:
the set of actions is a multivariate set, for instance [0, 1] × [0, 1]
the tangent function is a function of several variables
5. Multivariate optimisation ECON10005 3 / 13
Stationary point
A stationary point should verify:
∂f
= 0,
∂x1
···
∂f
=0
∂xn
for all choice variables: (x1 , · · · , xn )
5. Multivariate optimisation ECON10005 4 / 13
Finding a maximum
1. We find the stationary points of f :
∂f
∂x1 = 0
.. First-order conditions (FOCs)
.
∂f
∂xn = 0
2. We evaluate f at the boundaries of S (finite or limit at infinity).
3. Among these candidate solutions, the action (x1 , · · · , xn ) providing the highest
value is the maximum.
5. Multivariate optimisation ECON10005 5 / 13
Concavity
If f is concave over S,
a stationary point (if it exists!)
will always be a maximum over S.
5. Multivariate optimisation ECON10005 6 / 13
Convexity
If f is convex over S,
a stationary point (if it exists!)
will always be a minimum over S.
5. Multivariate optimisation ECON10005 7 / 13
Concavity with two variables
Concavity with one variable: f is concave over S if f ′′ (x) ≤ 0 for all x ∈ S
Concavity/ convexity is more difficult to establish with many variables
Concavity with two variables
f is concave over S if
∂2f
1. ≤0
∂x2
∂2f
2. ≤0
∂y 2
2
∂2f ∂2f ∂2f
3. − ≥0
∂x2 ∂y 2 ∂x∂y
5. Multivariate optimisation ECON10005 8 / 13
Convexity with two variables
Convexity with one variable: f is convex over S if f ′′ (x) ≥ 0 for all x ∈ S
Convexity with two variables
f is convex over S if
∂2f
1. ≥0
∂x2
∂2f
2. ≥0
∂y 2
2
∂2f ∂2f ∂2f
3. − ≥0
∂x2 ∂y 2 ∂x∂y
5. Multivariate optimisation ECON10005 9 / 13
Why do we need condition 3?
Example
f (x, y) = x2 + y 2 − 3xy
5. Multivariate optimisation ECON10005 10 / 13
Why do we need condition 3?
f (x, y) = x2 + y 2 − 3xy
5. Multivariate optimisation ECON10005 11 / 13
Economic Example: Profit maximisation
Consider an entrepreneur (Tobacco Yoko) producing tobacco with inputs labour L
and capital K:
1 1
Y = F (K, L) = K 2 L 4
We assume that markets are competitive: the entrepreneur takes the factor prices
r = 1.2 and w = 0.6 as given.
The unit price of tobacco is P = 12
5. Multivariate optimisation ECON10005 12 / 13
Reading
Sydsaeter et al. (2016). Essential mathematics for economic analysis, ch. 13.
5. Multivariate optimisation ECON10005 13 / 13
