Optimization in Economic Theory

PD Dr. Johannes Paha

University of Hohenheim

winter term 2021/22

[Link]@[Link] Optimization in Economic Theory (winter term 2021/22) 1 / 45

 Chapter 8 – Lagrange Multiplier & Envelope Theorem

Chapter 8
Lagrange Multiplier & Envelope Theorem

Dec 14 and 21, 2021

Optimization in Economic Theory

PD Dr. Johannes Paha
winter term 2021/22

[Link]@[Link] Optimization in Economic Theory (winter term 2021/22) 2 / 45

 Chapter 8 – Lagrange Multiplier & Envelope Theorem Double click here for audio contents

Overview – Comparative Statics

This lecture provides a closer look at the Lagrange parameter, it introduces max’m
value fn’s, and it derives the Envelope Theorem.
The Lagrange multiplier and the maximum value
Shadow prices
Maximum value functions: Unconstrained optimization
Concluding remarks
Background reading: Simon and Blume (1994, ch. 19)

[Link]@[Link] Optimization in Economic Theory (winter term 2021/22) 3 / 45

 Chapter 8 – Lagrange Multiplier & Envelope Theorem Double click here for audio contents

Learning goals
Students
Can compute the effects of marginal changes in the constraints
on the value of objective function
Can interpret Lagrange multipliers as shadow prices
Can explain the concept of a maximum value function
and the envelope theorem.
Can apply the Envelope Theorem
to the unconstrained maximization problem,
to compute the effects of marginal parameter changes on the value of
the objective function, and
to sit’ns where some of the choice variables are fixed in the short-run.
Can compute the effects of marginal parameters that affect both the objective
function and the constraint on the value of the objective fn.

[Link]@[Link] Optimization in Economic Theory (winter term 2021/22) 4 / 45

 Chapter 8 – Lagrange Multiplier & Envelope Theorem Double click here for audio contents

Shadow prices – The Lagrange multiplier

Recap: Interpretation of the Lagrange parameter

Consider a maximization problem with 2 choice variables
and 1 constraint
We have argued that the Lagrange multiplier λ can be interpreted as the change in
the value of objective function dν induced by relaxing the constraint by one unit
dc (if that unit is very small)

dν
dν = λdc ⇔ =λ
dc
In the proof of this claim, we used the total differential of the objective function
and the constraint.

[Link]@[Link] Optimization in Economic Theory (winter term 2021/22) 5 / 45

 Chapter 8 – Lagrange Multiplier & Envelope Theorem Double click here for audio contents

Shadow prices – The Lagrange multiplier

Generalization to n choice variables and m equality constraints

This generalizes to more variables and more constraints:
dν = F (x̄ + dx̄) − F (x̄) = Fx (x̄)dx̄
= λHx (x̄)dx̄ = λ [H(x̄ + dx̄) − H(x̄)]
= λ [c + dc − c] = λdc.
Dimensions
λ is an m-dimensional row vector
Hx (x̄) is an (m × n) matrix
dx̄ is an n-dimensional column vector
The final result is the product of the row vector λ and the column vector dc of
equal dimensions m such that it is a scalar:
X
λdc = λi dci .
i

[Link]@[Link] Optimization in Economic Theory (winter term 2021/22) 6 / 45

 Chapter 8 – Lagrange Multiplier & Envelope Theorem Double click here for audio contents

Shadow prices – The Lagrange multiplier

Extension to inequality constraints

Consider an increase in ci by dci while constraint i is slack in the initial optimum,
i.e. H i (x̄) < ci .
How will ν change? If constraint i is not binding,
the corresponding Lagrange multiplier is λi = 0
Applying the formula above gives dν = 0
This is exactly what we would expect:
If constraint i is slack, i.e. if it is optimal not to exhaust constraint i anyway,
relaxing this constraint will change neither the optimum choice nor the maximum
attainable value v .

[Link]@[Link] Optimization in Economic Theory (winter term 2021/22) 7 / 45

 Chapter 8 – Lagrange Multiplier & Envelope Theorem Double click here for audio contents

Shadow prices – Example: Sales maximization

Consider a firm that has a stock X > 0 of a good to sell on the market at price
p > 0.
The firm wants to maximize sales
subject to the constraints x ≥ 0 and x ≤ X .
The corresponding Lagrangian is
L(x , λ) = p · x + λ(X − x )
The first order conditions are
Lx = p − λ ≤ 0, x ≥ 0, (p − λ)x = 0
Lλ = X − x ≥ 0, λ ≥ 0, λ(X − x ) = 0

[Link]@[Link] Optimization in Economic Theory (winter term 2021/22) 8 / 45

 Chapter 8 – Lagrange Multiplier & Envelope Theorem Double click here for audio contents

Shadow prices – Example: Sales maximization

We have to consider 21+1 = 4 cases:

1 λ=0 and x =0
2 λ=0 and x >0
3 λ>0 and x =0
4 λ>0 and x >0
Case 1
The constraint is slack (λ = 0) and x = 0
Lx implies p ≤ 0, which is a contradiction to the assumption p > 0.
Case 2
The constraint is slack (λ = 0) and x > 0
Lx implies p = 0, which is a contradiction to the assumption p > 0.

[Link]@[Link] Optimization in Economic Theory (winter term 2021/22) 9 / 45

 Chapter 8 – Lagrange Multiplier & Envelope Theorem Double click here for audio contents

Shadow prices – Example: Sales maximization

Case 3
The constraint is binding (λ > 0) and x = 0
Lλ implies x = X > 0. Thus, there is a contradiction to the
assumption of a positive stock of the good.
Case 4
The constraint is binding (λ > 0) and x > 0
Lλ implies x = X > 0, which is consistent with a positive stock of the
good.
Lx implies λ = p.
Interpretation of this solution
The value of selling an extra unit (which is made possible by relaxing
the constraint) is the price of the good p.
Thus, the Lagrange multiplier reflects the price on the
corresponding constraint.

[Link]@[Link] Optimization in Economic Theory (winter term 2021/22) 10 / 45

 Chapter 8 – Lagrange Multiplier & Envelope Theorem Double click here for audio contents

Shadow prices – Units

The objective function can be in one unit,

while the constraints can be in different units
The Lagrange multiplier (the shadow price) is expressed in the unit of the objective
function per unit of the constraint
In the Lagrange function, the Lagrange multiplier turns the units of the constraint
into the units of the objective function so that the Lagrangian is in just one unit
In general, expressing the constraint in a different way changes the shadow price
and its interpretation

[Link]@[Link] Optimization in Economic Theory (winter term 2021/22) 11 / 45

 Chapter 8 – Lagrange Multiplier & Envelope Theorem Double click here for audio contents

Shadow prices – Units

Consider the utility maximization problem

The objective function is in units of utility
The budget constraint is expressed in monetary units
The Lagrange multiplier (the shadow price) is expressed in units of utility per
monetary unit
The inverse of the Lagrange multiplier is the monetary price one has to pay in
order to increase utility by one unit

[Link]@[Link] Optimization in Economic Theory (winter term 2021/22) 12 / 45

 Chapter 8 – Lagrange Multiplier & Envelope Theorem Double click here for audio contents

Shadow prices – Inequality constraints

Consider an inequality constraint H(x ) ≤ c

The first order conditions with respect to the corresponding Lagrange multiplier
are:
Lλ = c − H(x ) ≥ 0, λ ≥ 0 with complementary slackness

Note that Lλ ≥ 0 implies H(x ) ≤ c.

If a resource is not fully used (H(x ) < c), then its shadow price is zero (λ = 0).
A resource with a positive shadow price (λ > 0) has to be fully used (H(x ) = c)

[Link]@[Link] Optimization in Economic Theory (winter term 2021/22) 13 / 45

 Chapter 8 – Lagrange Multiplier & Envelope Theorem Double click here for audio contents

Maximum value functions – Unconstrained optimization

So far
Focus on the interpretation of the Lagrange parameter
That means, analyze changes in the maximum attainable value of the
objective function when the right hand side of the constraints are
altered slightly
But
Also the objective function depends on parameters
How do changes of the parameters of the objective function affect its
maximum attainable value?

[Link]@[Link] Optimization in Economic Theory (winter term 2021/22) 14 / 45

 Chapter 8 – Lagrange Multiplier & Envelope Theorem Double click here for audio contents

Maximum value functions – Unconstrained optimization

We will denote the vector of K parameters by θ

To make the dependency on the parameters explicit, we rewrite the objective
function as F (x, θ) where x is an n-dimensional vector.
Suppose that the decision problem does not involve any constraints.
Let x̄ denote the (unconstrained) maximizer of F (x, θ)
Consider changes in the parameters from θ to θ + dθ
Two effects on the value of the objective function
1 Since θ is an argument of the objective function, a change in θ has a
direct impact on F (x, θ).
2 The optimal choice x̄ depends on θ. A change in θ thus induces new
optimal values which have an impact on F (x, θ) as well.

[Link]@[Link] Optimization in Economic Theory (winter term 2021/22) 15 / 45

 Chapter 8 – Lagrange Multiplier & Envelope Theorem Double click here for audio contents

Maximum value functions – Formalization

Notation: Write Fxj (x, θ) for the partial derivative of F (x, θ) with respect to xj
and Fθk (x, θ) for the partial derivative with respect to θk , k = 1, . . . , K .
The first order Taylor series approximation of a change in the value of the
objective function is
n K
X X
dν = F (x̄ + dx̄, θ + dθ) − F (x̄, θ) = Fxj (x, θ)d x̄j + Fθk (x, θ)dθk
j=1 k=1
1 The first term captures changes in the objective function induced by
changes in optimal choices.
2 The second term captures the direct impact of the change in θ on the
objective function.

[Link]@[Link] Optimization in Economic Theory (winter term 2021/22) 16 / 45

 Chapter 8 – Lagrange Multiplier & Envelope Theorem Double click here for audio contents

Maximum value functions – Key insight

Important observation
x̄ has been chosen optimally given θ.
Evaluated at the optimum, the derivatives of the objective function with respect to
xj , j = 1, . . . , n are all zero: Fxj (x̄, θ) = 0.
Thus, small changes in the choice variables (here from x̄ to x̄ + dx̄) have
(approximately) no effect on the objective function such that:
K
X ∂ν ∂F (x, θ)
dν = Fθk (x, θ)dθk ⇔ = for all k
∂θk ∂θk
k=1

[Link]@[Link] Optimization in Economic Theory (winter term 2021/22) 17 / 45

 Chapter 8 – Lagrange Multiplier & Envelope Theorem Double click here for audio contents

Maximum value functions – Key insight

Summary
All that needs to be done is to evaluate the derivative of the maximum value
function with respect to θ at the original optimum x̄.
This is called the Envelope Theorem.

[Link]@[Link] Optimization in Economic Theory (winter term 2021/22) 18 / 45

 Chapter 8 – Lagrange Multiplier & Envelope Theorem Double click here for audio contents

Unconstrained optimization – Envelope Theorem

Consider an unconstrained optimization problem
with n choice variables.
The objective function contains K parameters collected in vector θ.
Let x̄ denote the maximand.
Evaluated at the optimum, the change in the value of objective fn dν induced by
changes in the parameters of the objective fn is:
n
X K
X
dν = Fxj (x̄, θ)d x̄j + Fθk (x̄, θ)dθk
j=1 k=1
K
X
= Fθk (x̄, θ)dθk
k=1
To find the first order change in the maximum value, we have to
compute the partial effect of the parameter change on the objective fn.
Key insight: We do not need to worry about the effect of the
simultaneous change in the optimal choice x̄.

[Link]@[Link] Optimization in Economic Theory (winter term 2021/22) 19 / 45

 Chapter 8 – Lagrange Multiplier & Envelope Theorem Double click here for audio contents

Unconstr’ed opt’n – Envelope Th’m (graphical illustration)

Consider two functions F (x̄1 , θ) and F (x̄2 , θ),

where x̄1 maximizes F (x, θ1 ) and x̄2 maximizes F (x, θ2 ).
Define the optimal value function
V (θ) = max F (x, θ)
x
The two functions V (θ) and F (x, θ) coincide in θ1 since x̄1 is the optimal choice
there.
For all other values of θ, x̄1 deviates from optimal choice
and thus F (x̄1 , θ) must lie below V (θ).
A similar argument applies to F (x̄2 , θ), which touches V (θ) at θ2
and lies below V (θ) otherwise.

[Link]@[Link] Optimization in Economic Theory (winter term 2021/22) 20 / 45

 Chapter 8 – Lagrange Multiplier & Envelope Theorem Double click here for audio contents

Unconstr’ed opt’n – Envelope Th’m (graphical illustration)

Now draw a whole family of functions F (x̄, θ) where in each function

x̄ is fixed to a value which is optimal for some θ.
Each member of this family would touch V (θ) once and would everywhere else be
strictly below it.

[Link]@[Link] Optimization in Economic Theory (winter term 2021/22) 21 / 45

 Chapter 8 – Lagrange Multiplier & Envelope Theorem Double click here for audio contents

Unconstr’ed opt’n – Envelope Th’m (graphical illustration)

[Link]@[Link] Optimization in Economic Theory (winter term 2021/22) 22 / 45

 Chapter 8 – Lagrange Multiplier & Envelope Theorem Double click here for audio contents

Unconstr’ed opt’n – Envelope Th’m (graphical illustration)

For any x̄, V (θ) is the envelope.

When the value of x̄ is chosen optimally given θ, V (θ) and F (x̄, θ) have the same
slope
dV (θ) dF (x̄ , θ)
=
dθ dθ
To compute dV (θ)/dθ, we can hold the value of x̄ fixed at the initial optimal
value.

[Link]@[Link] Optimization in Economic Theory (winter term 2021/22) 23 / 45

 Chapter 8 – Lagrange Multiplier & Envelope Theorem Double click here for audio contents

Example: Hotelling’s Lemma

Consider again a firm that employs labor as the only input and has the production
function y = ALα
Let p denote the price of output and w the cost of labor
Then, maximum profits are given by
π ≡ max {p · y (L) − w · L}
L
Applying the Envelope Theorem to the parameters p and w , we obtain
∂π ∂π
= ȳ and = −L̄
∂p ∂w
If the price of output increases by one unit, profits increase by ȳ units,
even if one takes possible changes in production into account.
If the cost of the input increases by one unit, profits decrease by L̄
units, even if one takes possible changes in production into account.

[Link]@[Link] Optimization in Economic Theory (winter term 2021/22) 24 / 45

 Chapter 8 – Lagrange Multiplier & Envelope Theorem Double click here for audio contents

Example: Hotelling’s Lemma

The previous insights are a direct consequence of

the Envelope Theorem.
Nevertheless, there is a special name to them:
Hotelling’s Lemma (named after Harold Hotelling).
Hotelling’s Lemma is widely used in the Theory of the Firm.
An important qualification is that the results only hold if the firm produces at the
optimum in the initial situation.

[Link]@[Link] Optimization in Economic Theory (winter term 2021/22) 25 / 45

 Chapter 8 – Lagrange Multiplier & Envelope Theorem Double click here for audio contents

Envelope Theorem for Unconstrained Opt’n Problems

Let F (x, θ) be a C 1 function of x ∈ Rn and a scalar θ. For each choice of the parameter
a, consider the unconstrained maximization problem
max F (x, θ) with respect to x.

Let x(θ) be a solution to this problem. Suppose that x(θ) is a C 1 function of θ. Then,
d
(x(θ), θ) · dx∂θ
i (θ)
P ∂F
dθ
F (x(θ), θ) = i ∂xi
+ ∂F
∂θ
(x(θ), θ) · 1
∂
= ∂θ
F (x(θ), θ)

∂F
because of ∂xi
(x(θ), θ) = 0 at the optimum.

(The partial derivative on the rhs is a lot easier to calculate than the total derivative on
the lhs because it can be determined even without calculating x(θ).)

[Link]@[Link] Optimization in Economic Theory (winter term 2021/22) 26 / 45

 Chapter 8 – Lagrange Multiplier & Envelope Theorem Double click here for audio contents

Constrained optimization

Suppose that parameters enter the objective function alone and are not part of the
constraints
Example 1: Preference parameters in optimal consumer choice model
(in the constraint, we have goods prices’, and income)

Example 2: Input prices in cost minimization problem to produce a

given amount of output (in the constraint, we have parameters from the production function)
If the solution has to satisfy m equality constraints, it can be found by solving the
Lagrangian m
X
L(x, λ, θ) = F (x, θ) + λi (ci − H i (x)).
i=1
First order conditions
∂L(x, λ, θ) ∂L(x, λ, θ)
=0 for all j and =0 for all i
∂xj ∂λi

[Link]@[Link] Optimization in Economic Theory (winter term 2021/22) 27 / 45

 Chapter 8 – Lagrange Multiplier & Envelope Theorem Double click here for audio contents

Constrained opt’n – Parameter changes

Again, suppose that θ changes to θ + dθ,

moving the optimal choice from x̄ to x̄ + dx̄
The new optimal choice x̄ + dx̄ still has to satisfy the constraints,
which are unaltered.
The change in θ has again two effects on the objective function
1 Direct effect (shift of the objective function)
2 Indirect effect through the change in θ to x̄ + dx̄.
First order Taylor approximation of change in value of objective fn
n
X K
X
dν = F (x̄ + dx̄, θ + dθ) − F (x̄, θ) = Fxj (x̄, θ)d x̄j + Fθk (x̄, θ)dθk
j=1 k=1
Important observation: With constrained m maximization, derivatives with respect
X
to xj are F xj (x̄, θ) = λi Hxi j (x̄).
i=1

[Link]@[Link] Optimization in Economic Theory (winter term 2021/22) 28 / 45

 Chapter 8 – Lagrange Multiplier & Envelope Theorem Double click here for audio contents

Constrained opt’n – Movements along the constraint

We obtain !
n
X n X
X m m
X n
X
Fxj (x̄, θ)d x̄j = λi Hxi j (x̄)d x̄j = λi Hxi j (x̄)d x̄j
j=1 j=1 i=1 i=1 j=1
m m
X X
λi H i (x̄ + dx̄) − H i (x̄) =


= λi · 0 = 0
i=1 i=1
which follows from the observation that x̄ and x̄ + dx̄ have to meet the constraint
Thus K
X ∂ν ∂F (x̄ , θ)
dν = Fθk (x̄ , θ)dθk ⇔
=
∂θk ∂θk
k=1
Summary: When parameters do not affect constraints, we can again ignore the
impact of changes in optimal choices on objective function.

[Link]@[Link] Optimization in Economic Theory (winter term 2021/22) 29 / 45

 Chapter 8 – Lagrange Multiplier & Envelope Theorem Double click here for audio contents

Constrained opt’n – Discussion

The finding is a direct implication of the no-arbitrage argument.

Since the constraints are unaffected by changes in θ, the vector dx̄ must be chosen
such that constraints do not change.
For example, if constraints increase in all xj , some elements of dx̄ must be
positive, some negative.
dx̄ reallocates some amount dci in the value of constraint function H i (c) away
from choice variables which are reduced and toward choice variables that are
increased
But: From the no-arbitrage argument, such a reallocation cannot have an effect on
the value of the objective function.

[Link]@[Link] Optimization in Economic Theory (winter term 2021/22) 30 / 45

 Chapter 8 – Lagrange Multiplier & Envelope Theorem Double click here for audio contents

Example: Cost minimization

Consider a cost minimization problem in which

x is the vector of inputs,
θ is the row vector of input prices.
H the production function,
c the required output level,
Then the producer minimizes θx if it maximizes −θx.
We have the following problem:
max F (x, θ) = −θx subject to H(x) = c.
Rewrite the resulting maximum as (−ν). Then from above we have
d(−ν) = Fθ (x̄, θ)dθ = −x̄dθ.

[Link]@[Link] Optimization in Economic Theory (winter term 2021/22) 31 / 45

 Chapter 8 – Lagrange Multiplier & Envelope Theorem Double click here for audio contents

Example: Cost minimization

This leads to the overall result dν = dθx̄.

When input prices change, the producer adjusts by adapting its optimal choice.
However, the term x̄ is still the original opt’l choice.
By contrast, if we differentiate ν = θx̄, we would get
dν = θdx̄ + dθx̄.
The first term is the value of the change in the input mix using original prices.
However, at those prices the original mix is chosen optimally.
The value of any change in the choice must be zero.
This means that only the second term is present and the first-order effect of the
parameter change on the cost is just the change in the cost at the original
optimum.

[Link]@[Link] Optimization in Economic Theory (winter term 2021/22) 32 / 45

 Chapter 8 – Lagrange Multiplier & Envelope Theorem Double click here for audio contents

Constrained opt’n – Parameters affecting all functions

Suppose now that both the objective function F

and the constraint H depend on θ.
In order to highlight the dependence of H on θ, we write
H i (x, θ) = ci ,
where ci is not part of θ.
The change in the value of the objective function still emerges as
n K
X X
dν = F (x̄ + dx̄, θ + dθ) − F (x̄, θ) = Fxj (x̄, θ)d x̄j + Fθk (x̄, θ)dθk
j=1 k=1

[Link]@[Link] Optimization in Economic Theory (winter term 2021/22) 33 / 45

 Chapter 8 – Lagrange Multiplier & Envelope Theorem Double click here for audio contents

Constrained opt’n – Parameters affecting all functions

But when a change in parameters θ affects the constraints, the first term
Pn
F (x̄, θ)d x̄j is no longer zero
j=1 xj

The reason is that a change in θ affects the constraint.

In addition to reallocating the choice variables along the initial constraints, the
decision maker can now increase all xj
(in cases where the change in θ relaxes the constraints).

[Link]@[Link] Optimization in Economic Theory (winter term 2021/22) 34 / 45

 Chapter 8 – Lagrange Multiplier & Envelope Theorem Double click here for audio contents

Constrained opt’n – Parameters affecting all functions

To calculate the effect, is is still useful to work with the FOCs

m
X
Fxj (x̄, θ) = λi Hxi j (x̄, θ)
i=1
Recall that we can rewrite this as !
n n m m n
X X X X X
Fxj (x̄, θ)d x̄ = λi Hxi j (x̄, θ)d x̄ = λi Hxi j (x̄, θ)d x̄j
j=1 j=1 i=1 i=1 j=1
m
X
λi H i (x̄ + dx̄, θ) − H i (x̄, θ)


=
i=1
Key difference: If θ affects the constraint, dx̄ can be chosen such that
H i (x̄ + dx̄, θ) ̸= H i (x̄, θ)

[Link]@[Link] Optimization in Economic Theory (winter term 2021/22) 35 / 45

 Chapter 8 – Lagrange Multiplier & Envelope Theorem Double click here for audio contents

Constrained opt’n – Parameters affecting all functions

Approximation of the change in the constraint

K
X
dH i = H i (x̄ + dx̄, θ) − H i (x̄, θ) = − Hθi k (x̄, θ)dθk ,
k=1
which follows from totally differentiating H(x, θ) = c with respect to x and θ

Hx (x, θ)dx + Hθ (x, θ)dθ = 0

Inserting delivers !
m K
X X
Fxj (x̄, θ) = − λi Hθi k (x̄, θ)dθk
i=1 k=1
The change in value of the objective function !
K
X m
X K
X
dν = Fθk (x̄, θ)dθk − λi Hθi k (x̄, θ)dθk
k=1 i=1 k=1

[Link]@[Link] Optimization in Economic Theory (winter term 2021/22) 36 / 45

 Chapter 8 – Lagrange Multiplier & Envelope Theorem Double click here for audio contents

Constrained opt’n – Parameters affecting all functions

The expression on the previous slide is a first order Taylor approxi- mation of the
change in the Lagrange function induced by change in θ
This implies
K
X
dν = Lθk (x̄, λ, θ)dθk
k=1

Intuitively, a change in θ now has

PK
1 Direct effect on F measured by k=1 Fθk (x̄, θ)dθk
2 Effect on constraints: depending on whether this effect is positive or
negative, a change in θ relaxes or tightens constraint i by
PK i
k=1 Hθk (x̄, θ)dθk . Since each unit of relaxation of constraint i
delivers λi additional value of the objective function, the result follows.

[Link]@[Link] Optimization in Economic Theory (winter term 2021/22) 37 / 45

 Chapter 8 – Lagrange Multiplier & Envelope Theorem Double click here for audio contents

Constrained opt’n – Some choice variables fixed

Summary
When parameters θ in the optimization problem change, so do the
optimal choice variables x and the maximum value ν.
The first order effect on ν can be calculated by holding x fixed at the
initial optimum x̄ and finding the partial effect of θ on ν.
If the parameters affect the constraints, we must include the
contribution of the equivalent reduction in the right-hand-side
magnitudes of c, but we can still ignore the change in x.
Generalization
What happens when only some components of x adjust to the new
optimum levels, while others must be kept fixed at the original level?
Example – Distinction between short- and long-run perspectives:
Only some variables can be adjusted in the short-run, while others can
only be adjusted in the long-run
[Link]@[Link] Optimization in Economic Theory (winter term 2021/22) 38 / 45
 Chapter 8 – Lagrange Multiplier & Envelope Theorem Double click here for audio contents

Constrained opt’n – Formalization

Separate the choice vector x into two subvectors y and z

In the short-run, only y is allowed to vary
In the long-run, both y and z are allowed to vary
The maximization problem then emerges as
maximize F (y, z, θ) subject to H(y, z, θ) = 0
In the long-run, y and z are the choice vectors
In the short-run, only y is the choice vector
The number of constraints has to be smaller than the number of the choice
variables in the short-run scenario

[Link]@[Link] Optimization in Economic Theory (winter term 2021/22) 39 / 45

 Chapter 8 – Lagrange Multiplier & Envelope Theorem Double click here for audio contents

Constrained opt’n – Formalization

The long-run choices and the maximum value can be written as functions of the
parameters θ y = Y (θ), z = Z (θ), ν = V (θ)
In the short-run, z should be treated as a parameter, and the short-run choices and
the short-run maximum values emerges as
y = Y (z, θ), ν = V (z, θ)
The definition of optimization implies
V (θ) ≥ V (z, θ) for all (z, θ),
with equality if z = Z (θ) happens to be optimal long-run choice

[Link]@[Link] Optimization in Economic Theory (winter term 2021/22) 40 / 45

 Chapter 8 – Lagrange Multiplier & Envelope Theorem Double click here for audio contents

Constrained opt’n – Application of the Envelope Theorem

Recall that
V (θ) ≥ V (z, θ) for all (z, θ),
with equality if z = Z (θ) happens to be optimal long-run choice.
The structure is comparable to the one that emerged in the standard application of
the Envelope Theorem.
The graph of V (θ) is the upper envelope of the curves showing V (z, θ)
as functions of θ for the whole range of possible values of z
The Envelope Theorem then implies
V ′ (θ) = Vθ (Z (θ), θ),
where the right hand side is the partial derivative of the short-run
optimum value function V (z, θ) when holding the first argument z
fixed, but evaluated at the point z = Z (θ)

[Link]@[Link] Optimization in Economic Theory (winter term 2021/22) 41 / 45

 Chapter 8 – Lagrange Multiplier & Envelope Theorem Double click here for audio contents

Summary: The Envelope Theorem

Effects of parameter changes (objective function) on the optimal value of the
objective function
Direct effect on the objective function
Indirect effect through adjustment in optimal choices
Envelope Theorem
Parameters of the objective function: Evaluate the partial
derivatives of the optimum value function with respect to the
parameter at the original optimum x̄.
K
X ∂ν ∂F (x̄, θ)
dν = Fθk (x̄, θ)dθk ⇔ = for all k
∂θk ∂θk
k=1
Parameters of the constraint and the objective function:
A change in a parameter related to a constraint works like relaxing
(or tightening) this constraint.
K
X ∂ν ∂L(x̄, θ)
dν = Lθk (x̄, λ, θ)dθk ⇔ = for all k
∂θk ∂θk
k=1
[Link]@[Link] Optimization in Economic Theory (winter term 2021/22) 42 / 45
 Chapter 8 – Lagrange Multiplier & Envelope Theorem Double click here for audio contents

Envelope Theorem for Constrained Opt’n Problems

Let F , H 1 , . . . , H K : Rn × R1 → R1 be C 1 functions. Let x(θ) = (x 1 (θ), . . . , x n (θ))

denote the solution of the problem of maximizing x → F (x, θ) on the constraint set

H 1 (x, θ) = 0, . . . , H K (x, θ) = 0,

for any fixed choice of the parameter θ. Suppose that x(θ) and the Lagrange multipliers
λi (θ, . . . , λi (θ)) are C 1 functions of θ and that NDCQ holds. Then,

d ∂L
F (x(θ), θ) = (x(θ), λ(θ), θ),
dθ ∂θ
where L is the natural Lagrangian for this problem.

[Link]@[Link] Optimization in Economic Theory (winter term 2021/22) 43 / 45

 Chapter 8 – Lagrange Multiplier & Envelope Theorem Double click here for audio contents

Concluding remarks

The observation that the Lagrange parameter represents the effect of a marginal
change in the constant c of the constraint generalizes to the case with n variables
and m < n constraints.
The Lagrange multiplier can be interpreted as a shadow price that puts a price on
any constraint.
With unconstrained optimization, we can ignore adjustments in the choice
variables when we compute the effects of parameter changes on the optimal value
of the objective function (Envelope Theorem).

[Link]@[Link] Optimization in Economic Theory (winter term 2021/22) 44 / 45

 Chapter 8 – Lagrange Multiplier & Envelope Theorem Double click here for audio contents

Concluding remarks

We have explored the effects of parameter changes on optimal choice and the
maximum value of the objective function
in unconstrained problems,
in constrained problems,
when parameters that only affect the objective function change, and
when parameters that affect both the objective function and the
constraints change.
We have seen that in all but the last case the Envelope Theorem heavily simplifies
the analysis.
We have related the latter to changes in the constant of the constraint.

[Link]@[Link] Optimization in Economic Theory (winter term 2021/22) 45 / 45