ECON-803 Micro I

September 8, 2025
Utility Maximization
September 8, 2025
Outline of our unit on consumer theory

1. A consumer’s preferences
• Preferences
• Representing preferences with a utility function
2. A consumer’s choices
• [TODAY] Utility maximization subject to a budget constraint 
• A consumer’s demand functions
3. Understanding demand functions
Last time

Fix a set of choices , e.g. .

A consumer’s preferences express her ranking of the choices in .

A utility function represents her preferences if the numbers it assigns to

the various choices are consistent with her ranking. (Higher numbers for
more-preferred choices.)
Budget Sets
• Typically, only a subset of the options in are feasible choices for the consumer. We call this
subset her budget set.
• Usually focus on Walrasian budget sets:

: vector of prices
: vector of quantities of each good
: income

• Example if there are two goods ( ) ?

• Hidden assumptions?
• (What does it mean to treat the per-unit prices as fixed parameters?)
• Once-and-for-all decision.
• (No notion yet of time and now-vs.-later tradeoffs)
• Treating income as a fixed parameter. Later, we could model where income comes from.
• E.g., consumer is endowed with quantities of each good: .
• Her income is the market value of that endowment: . Budget constraint .
Utility Maximization Problem

For the typical case where , and a choice is a consumption bundle over
goods, the consumer’s decision problem is:

• The vector that solves this problem is her vector of demand

functions:

• The utility she achieves from her best (affordable) bundle is her
indirect utility function:
UMP Example: CES utility with

Choice set , utility .

UMP:

𝑥2
Two alternative approaches to this UMP:
• Tangency + BC
• If confident that the picture qualitatively
reflects the situation, then solution pinned
down by
1. Tangency condition: MRS = price ratio.
𝒑 ⋅ 𝒙=𝑌
𝑥1
2. BC: bundle is on budget line .

• Lagrangian approach
 Tangency + BC Approach

Optional first step: work with instead. (Why is this OK?)

Marginal utilities:

Slope of indifference curves:

Slope of budget line:

Tangency condition:

Equivalent to:
is equalized across all goods. (This version applies with more than two goods. Interpretation?)
Side note: CES tangency condition

In our example, the tangency condition implies that an optimal choice

satisfies:

• This power-law relationship between the consumption ratio and price

ratio is, in some sense, the defining feature of CES utility.
• The exponent is this consumer’s (constant) elasticity of substitution, .
(More on elasticities of substitution later…)
Back to the UMP: conditions for an optimal choice

1. (Tangency)
2. (Budget constraint)
Plug (1) into (2) to solve for demand functions and .

(The bits that change are highlighted in blue and red.)

Smell test:
Do our solutions behave like we expect demand functions to behave?

Our solution for demand for good 1:

What is the sign of the change in demand for good 1 …

If income rises? ( )

If its own price rises? ( )

If prices and income all double (or triple)?* ( )

(*Not a coincidence – we’ll see that demand function are homogeneous of degree 0 in
prices and income.)
Solving the same UMP by the Lagrangian
approach
Actually, slight change to goods (since is where the Lagrangian
approach shines).

(Let’s work with the untransformed , just to see that this works too.)
Set up the Lagrangian for this constrained maximization problem:

Solve

(Put a pin in these thoughts for later: How do we know whether to insert the constraint with a + or a ?
And what about non-negativity of ?)
Lagrangian FOCs

We have choice variables: and .

The first FOCs correspond to the tangency condition from earlier:

And the final FOC stipulates that consumption is on the budget line:
FOCs: MU/$ is equalized across goods

The first FOCs can be summarized as:

This gives us one interpretation of the multiplier .

Using the FOCs to solve for
(Same solution as earlier, adjusting for .)
Write , .
• Not necessary. Just helps me be tidy with notation and the Chain Rule.
Marginal utility for good :

FOC says for any

Rearrange as (same as earlier!):

Last step: plug into the BC
Conceptually straightforward, but sometimes is algebraically messy. (Exploiting any
symmetry present in the problem can help.) E.g.

So

Plug into BC:

Or – not obvious why this is a useful piece of shorthand, but be patient – define by

So, demand for good is:

(And similarly for goods .) Interpretation of the terms in this expression?

Indirect Utility
We’ve found the optimal choices this consumer makes when facing the budget :

The utility she realizes at these optimal choices is called her indirect utility function.

In our example (this is not obvious, need to plug in and simplify):

Remember the Lagrangian:

At the consumer’s optimal bundle, , and so the red term vanishes. So it is also true that

So, demand for good is:

Economic interpretations of the Lagrange multiplier

Interpretation 1:
We already saw that at the optimal consumption bundle,

Interpretation 2:
This will be hand-wavy until we’ve covered the Envelope Theorem.

So a small increase in income improves the consumer’s optimized utility

by
The utility maximization problem – technical
considerations
• The FOCs are necessary conditions for an interior maximum.
• That is, if the true optimal bundle involves strictly positive quantities of all
goods (, etc.), then must satisfy the FOCs.
• What about sufficient conditions for a maximum?
• Can we be sure that [ ] even has a solution?
• If we find a solution to the FOCs, when does that guarantee that our solution
is a local max (not a minimum – oops!)? And when does this guarantee our
solution is a global max (which is our goal, after all)?
• And how do we recognize and handle “messier” cases?
• (Corner solutions, non-differentiability, cases where the FOCs might flag a
local min, etc.)
Does a solution to the UMP exist?

Result A continuous function on a compact set attains a maximum.

• If all prices are strictly positive, the Walrasian budget set

is compact. So as long as is continuous, the UMP does have a

solution.
• What goes wrong if one or more prices is equal to zero?
Second-Order Conditions for UMP

Mainly:
• At least once or twice: read/reread technical details in text until
comprehension.
• After that: look up the technical details if and when you need them.

Here:
• Example on the board where SOCs are violated.
• Derive second-derivative condition in terms of the Hessian matrix.
Recognizing and dealing with messier cases

• Let’s start by introducing some of the “usual suspect” utility

functions.
• Some of these are very well-behaved.
• The others give a good introduction to dealing with special cases.
The usual suspects: the CES family

General form:

Elasticity of substitution between goods is (informally):

(change in relative consumption induced by a change in relative prices)

More formally, we can write

Typically, this elasticity depends on the bundle where we evaluate it.

The usual suspects: the CES family

General form:

But for CES utility, is constant!

The tangency condition is:

So,
(regardless of which consumption bundle we evaluate at)
CES family members

Leontief* preferences:
• a.k.a. perfect complements, or fixed proportions
• Limit of CES as .
• (Write the CES parameters as and then look at the limit of the tangency condition.)

Cobb-Douglas preferences:
• Limit of CES as .

Linear* preferences:
• a.k.a. perfect substitutes
• Special case of CES when

CES utility (including these special cases) is homothetic. (Implications?)

*These are special cases when solving a utility-maximization problem. Don’t start differentiating
blindly!
The usual suspects: Quasilinear utility

General form:

Example. (Illustrates how to recognize a corner solution.)

Solve the two-good UMP , where

Method 1: Ask forgiveness, not permission.

• Ignore the non-negativity constraints for now.
• If and when our “solution” violates them, patch the solution up.
Method 2: Include constraints from the start.
Method 1

FOCs

Proceed as we normally would.

• Solve (2) for , subst. into (1) to get: .
• Plug this into BC to get .
Pause, assess the candidate solution with common sense.
• If , candidate sol’n says . Oops!
• So in the true sol’n, must bind.
• Alternatively, if , our candidate solution satisfies the non-negativity constraints we
ignored. So it’s correct, and we’re done!
Result

If you have some economic intuition about which of the extra

constraints are likely to bind or be slack, Method 1 is almost always
simpler.
Method 2

(With 20-20 hindsight, I’ve still omitted the constraint. We now know it won’t
bind, and it would just clutter the analysis.
The FOCs

So the FOCs “try” to equate MU/$.

• If this is possible in the feasible ( range, that’s the solution. Then is slack, and
so .
• But if MU/$ is strictly higher for good 1 – even when consumption is tilted to
all good 1, no good 2 – then the best feasible option is the corner (and the
term corrects the FOCs for the degree to which MU/$ isn’t equalized.
Sidebar: declining MRS vs. declining MU
The quasilinear example CES, (for example)

Slope
𝑥2 𝑥2 Slope

Slope Slope

𝑥1 𝑥1
Sidebar: declining MRS vs. declining MU
The quasilinear example CES, (for example)

Slope
𝑥2 𝑥2 Slope

Slope Slope
𝒑 ⋅ 𝒙=𝑌
𝑥1 𝑥1

Never consume , even if is high (but finite).

 Declining MRS declining MU
The important implications here about consumer choice depend on the
MRS declining. (And whether it asymptotes to and in the “corners.”)
Remember, . So declining MU implies declining MRS, but the reverse is
not true.
Slope
𝑥2 𝑥2 Slope

Slope Slope
𝒑 ⋅ 𝒙=𝑌
𝑥1 𝑥1
More importantly, the same preferences can have declining, constant, or
increasing MU, depending on how we normalize utility. E.g., , , or . So
focusing on declining MU is a red herring – pay attention to what the
Special cases: Leontief utility

Example:
Don’t start differentiating.
Any bundle where is suboptimal.
• Why?
So, set , plug into BC to get the optimal bundle.
Expenditure Minimization Problem & Duality

• The utility-maximization problem takes consumer’s budget set as

given and solves for the highest affordable indifference curve.
• The expenditure minimization problem reverses this: fix a target
indifference curve and solve for the cheapest bundle that attains it.
• Duality theory: the UMP and EMP are different facets of the same
problem. We can use one to shed light on the other.
• Envelope Theorem: key tool for understanding how UMP and EMP are
affected by price changes. (Also important in many other settings.) 
• Income and Substitution Effects. A clean description of these is one of
the major payoffs from fleshing out the concepts above.
Utility maximization vs. Expenditure
minimization
UMP: EMP:

Bundle solving the UMP is Bundle solving the EMP is

uncompensated, or compensated, or “Hicksian”
“Marshallian” demand: demand:
EMP: Cobb-Douglas example

Solve

Hicksian demand: ,
Expenditure function:
Note in passing the special feature of C-D preferences: regardless of
prices, fraction of spending on each good is:
Duality
[Subject to some regularity
conditions],
If bundle solves:
Maximize , subject to the 𝑥2 𝑢 ( 𝒙 )=𝑢
red budget set,
𝑍
Then also solves
Minimize , subject to 𝒑 ⋅ 𝒙=𝑌
reaching the blue 𝑥1
indifference curve.
Duality Identities
[Notice that all of these are
for a fixed set of prices.]

𝑥2 𝑢 ( 𝒙 )=𝑢

In words: 𝑍
If is the best utility
attainable with income , 𝒑 ⋅ 𝒙=𝑌
then the cheapest way to 𝑥1
achieve utility requires
spending dollars.
Duality Identities
[Notice that all of these are
for a fixed set of prices.]

𝑥2 𝑢 ( 𝒙 )=𝑢

𝒑 ⋅ 𝒙=𝑌
𝑥1
Duality Identities for Marshallian vs. Hicksian
demand
[Notice that all of these are
for a fixed set of prices.]

𝑥2 𝑢 ( 𝒙 )=𝑢

𝒑 ⋅ 𝒙=𝑌
𝑥1
Duality example: revisit the Cobb-Douglas eg from a few
slides up

We solved the EMP directly to get .

Alternatively, we could have found by:
1. Solve UMP:
 FOCs ,
 Plug into to get
2. Set and solve for in terms of . (Same function as above.)
Envelope Theorem & its uses in consumer theory

We’re often interested in how the result of an optimization varies with

parameters. E.g.

UMP:
• How does optimized utility vary with income? (What is ?)

EMP:
• How is the expenditure required to achieve affected by a price increase for
good 1? (What is ?)
Unconstrained version of Envelope Theorem

Here’s a generalized maximization problem with choice variable and

parameter :

Observe that

So a change in affects the optimized value in (potentially) two ways:

• Direct effect: .
• Indirect effect: affects optimal choice , change in affects .
Envelope Theorem: Indirect effect vanishes!
Example:

Envelope Theorem: Proof

Intuition: Change in does induce change in choice . But, in neighborhood of optimal , changing
doesn’t affect .
Version of Envelope Theorem for constrained
optimization

Lagrangian

Suppose optimal solution satisfies Lagrangian FOCs.

Then
Applications of Envelope Theorem to Consumer
Theory
• Expenditure function
• Shephard’s Lemma 
• (Links expenditure, price changes, Hicksian demand.)
• Indirect utility function
• Interpretation of Lagrange multiplier on budget constraint.
• Roy’s Identity. (Links indirect utility, price & income changes, Marshallian
demand.)
• Income and substitution effects
• The Slutsky equation decomposes into income, substitution effects.
• Relies on Envelope Theorem + duality relations between and between .
Shepard’s Lemma (involves expenditure function and Hicksian
demand)

Proof: [Apply Env. Thm]

• (We might do this on the board.)
𝑥2 𝑢 ( 𝒙 )=𝑢
Intuition:
For example, in diagram to the right, for small ,
why doesn’t the induced substitution from good 1 𝒑 ⋅ 𝒙=𝑐𝑜𝑛𝑠𝑡𝑎𝑛𝑡 𝒉(𝒑, 𝑢)
 good 2 affect total spending?
h 2 (𝒑 , 𝑢)
Can we (rigorously) confirm that the indirect
effects vanish without invoking the Envelope
Theorem?
h 1(𝒑 ,𝑢) 𝑥1
• Yes. Take derivative, applying the Chain Rule.
• Use FOCs to convert the indirect-effect term
from a change in $ to a change in .
• But this is zero, since we are evaluating along
the indifference curve.
Roy’s Identity (involves indirect utility and Marshallian
demand)

How (and why?) do we come to this expression?

1. Inspired by Shephard’s Lemma, look for a similar result with indirect utility.
2. Good news: does involve . But bad news: it also involves .
3. Looking for a way to get rid of and isolate leads us to .