Bilkent University

Econ 101 - Fall 2022

Chapter 2: Consumer Theory

A. Arda Gitmez Nuh Aygün Dalkıran

September 21, 2022

Contents
1 A Very Brief Introduction 2

2 The Model 2
2.1 The Constraint . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2 Preferences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2.1 Indifference Curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2.2 Marginal Rate of Substitution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

3 Optimal Bundle 7

4 Changes in Parameters 11
4.1 What If the Income Changes? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
4.2 What If the Price of a Good Changes? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

A Proofs 20

B Optimal Bundle for Some Famous Preferences 21

B.1 Perfect Substitutes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
B.2 Quasi-linear Preferences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
B.3 Cobb-Douglas Preferences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
B.4 Perfect Complements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

1
1 A Very Brief Introduction
This lecture introduces the first formal model of the decision made by an economic agent: a consumer. We
consider the simplest (and possibly the most widespread) form of economic interaction: the decision to buy
some goods and services. This is literally the model of a consumer who is deciding to buy some goods in a
supermarket. There are n different goods, which have their own prices. The consumer observes the prices
and decides how much to buy from each good.
As we discussed in the previous lecture, an economic agent has (i) a constraint and (ii) preferences. This
economic agent is no different: the consumer has limited income, denoted by a number I. This is the
consumer’s constraint: she cannot buy everything she wants. The consumer also has well-defined preferences
towards goods. We will discuss what “well-defined” means in a few pages.
Also, as discussed in the previous lecture, each economic model involves some simplifications. Here are a
couple of simplifications we assume throughout this document:
• This is a one-time interaction. The customer does not go to the store and say “let me buy this item
later”. There may be future considerations in her decision to buy some goods (i.e., if she is buying a
washing machine she realizes that she will most likely not need a washing machine in the near future).
Those are fine – the future considerations, concerns about the quality of the good, uncertainty about
its use, social concerns etc. are all captured by her preferences.
• The customer observes all the prices perfectly, and knows her income. She can easily calculate the
money required to buy a given group of items.
• The customer knows her preferences, and she acts according to her preferences. That is, she is rational.
At the risk of being repetitive: do these assumptions sometimes violated? Of course. But we are building a
benchmark here.

2 The Model
There is a single consumer and n different goods. The consumer decides how much to buy of each good, i.e.
she chooses quantities.
Here is the notation we will use:
• i: Will be used to denote a generic good. We will use natural numbers to denote goods and to index
them. Thus, the set of all goods will be denoted by {1, 2, . . . , n}. Here n denotes the nth good, and
also the total number of goods that is available for consumption.
• qi : Denotes the quantity of good i. The case where the consumer is considering consuming 5 kg of
good 2 will be represented with q2 = 5 kg. The quantity can be kilograms, grams, liters, numbers...
Whatever the denomination is, we will say it is a unit. Note that for any good i we must always have
qi ≥ 0. qi = 0 is allowed, i.e. the consumer may choose not to buy a good.
• (q1 , q2 , . . . , qn ): Denotes a consumption bundle (or simply a bundle). This is a list that represents
how much of each good a consumer is considering for consumption. As an example, consider a situation
where there are 4 possible goods that the consumer can consume (hence n = 4). The consumption
bundle (8, 2, 0, 12) represents the situation where the consumer is considering 8 units of good 1, 2 units
of good 2, none of good 3, and 12 units of good 4 for consumption.
• pi : Denotes the price of good i per unit. Therefore, if the consumer buys qi units of good i at price pi ,
she pays pi qi for that good.
• I: Denotes the income of the consumer (the total wealth of the consumer).

2
2.1 The Constraint
The constraint of the consumer specifies which bundles are affordable (i.e., feasible for the consumer) and
which bundles are not.
Definition 1. Given the prices p1 , p2 , . . . , pn of the goods and the income I of the consumer, a bundle
(q1 , q2 , . . . , qn ) is feasible if and only if
n
X
pi qi = p1 q1 + p2 q2 + · · · pn qn ≤ I .
i=1

The set of feasible bundles is also called the budget set. The set of feasible bundles that requires the use of
all the income, i.e., the bundles (q1 , q2 , . . . , qn ) such that p1 q1 + p2 q2 + · · · + pn qn = I are said to be on the
budget line (they constitute the budget line).
When there are two goods (n = 2), the set of feasible bundles given the prices p1 , p2 and income I are the
bundles (q1 , q2 ) that satisfy:
p1 q1 + p2 q2 ≤ I
They can be represented graphically with the orange region shown in Figure 1. The budget line is a line
with slope − pp12 .

q2

I
p2
Bu
dg
et
Li
ne

Budget
Set

I
q1
p1

Figure 1: A graphical representation of feasible bundles given prices p1 , p2 , and income I.

2.2 Preferences
The preferences of the consumer specify the consumer’s ranking between any two bundles q = (q1 , q2 , . . . , qn )
and q ′ = (q1′ , q2′ , . . . , qn′ ). The consumer may strictly prefer one bundle over the other, or may feel indifferent
between them. We capture the consumer’s preferences through a preference relation.
A preference relation is a relation that compares two pairs of bundles. It is defined on all pairs of
bundles, including (but not limited to) feasible bundles. Given two bundles q and q ′ , it contains three
possible scenarios.
• The situation where the consumer (strictly) prefers bundle q = (q1 , q2 , . . . , qn ) to another bundle
q ′ = (q1′ , q2′ , . . . , qn′ ) is represented by:
q P q′

3
 • If the consumer is indifferent between the bundles q and q ′ , then we will write:

q I q′

• If, for the consumer, bundle q is at least as good as bundle q ′ (i.e., the consumer prefers q to q ′ or
she is indifferent between q and q ′ ) we will write:

q R q′

Therefore, for any pair of bundles q and q ′ , we have:

q R q ′ ⇐⇒ (q P q ′ or q I q ′ )

Note that when q P q ′ , by definition, it is also true that q R q ′ . This is not really surprising: when a
consumer prefers q to q ′ , she also thinks q is at least as good as q ′ .
Similarly: when q I q ′ , by definition, it is also true that q R q ′ . This is not surprising either: when a
consumer is indifferent between q to q ′ , she also thinks q is at least as good as q ′ .
What I want to point out that two of these scenarios can be satisfied at the same time. This is just
like comparisons of real numbers. For the complete analogy: P is like the > sign, I is like the = sign,
and R is like the ≥ sign. Now, as we know, 5 > 3 and 5 ≥ 3 are both correct. Similarly, 4 = 4 and
4 ≥ 4 are both correct.
A “well-defined” preference relation satisfies the following three conditions:
1. For any pair of bundles q and q ′ ,

q R q ′ or q ′ R q.

This condition states that the consumer is able compare any pair of bundles. A relation that satisfies
this condition is said to be complete.
2. For any triple of bundles q, q ′ , and q ′′ ,

if q R q ′ and q ′ R q ′′ , then q R q ′′ .

That is, for the consumer, if q is at least as good as q ′ and q ′ is at least as good as q ′′ , then q should
be at least as good as q ′′ . A relation that satisfies this condition is said to be transitive.
3. For any pair of bundles q = (q1 , q2 , . . . , qn ) and q ′ = (q1′ , q2′ , . . . , qn′ ),

if qi < qi′ for all i ∈ {1, 2, . . . , n}, then q ′ P q.

This condition states that if the bundle q ′ has more of each good than bundle q has, then q ′ is preferred
to q. To put it simply, the consumer prefers more to less. A relation that satisfies this condition is
called monotonic.
Throughout this lecture, we assume that any preference relation is complete, transitive and monotonic, i.e.,
it is well-defined.

2.2.1 Indifference Curves

A graphical representation of preference relation can be obtained by drawing curves through bundles that the
consumer is indifferent among. Figure 2 gives information about a preference relation in which the consumer
is indifferent between the bundles q and q ′ (they are on the same curve).
The curve that passes through q is called indifference curve through q. Since q ′′ and q ′′′ are not on the
indifference curve through q, the consumer is not indifferent between q ′′ and q and also is not indifferent
between q ′′′ and q. Since q ′′ contains more of the (two) goods than q does, by the monotonicity of the
preference relation, q ′′ is preferred to q. Similarly, since the bundle q contains more of the goods than q ′′′

4
 q2

q0
q 00

q 000

q1

Figure 2: A graphical representation of a preference relation.

does, by the monotonicity of the preference relation, q is preferred to q ′′′ . Extending this argument, we can
conclude that any bundle that lies “above” (in the north east side of) the indifference curve through q is
preferred to q and q is preferred to any bundle that lies “below” (in the south west side of) the indifference
curve through q.
By now, you may have realized that it is possible to draw multiple indifference curves in the same graph.
Consider, for instance, the indifference curve passing through q ′′ . See Figure 3.
• By definition, the consumer is indifferent between any bundle on this “higher” indifference curve and
q ′′ . For instance, the consumer is indifferent between q ′′′′ and q ′′ :

q ′′′′ I q ′′

• Recall that, by monotonicity, the consumer prefers q ′′ to q:

q ′′ P q

• Once again, by definition, the consumer is indifferent between any bundle on the “lower” indifference
curve and q. For instance, the consumer is indifferent between q ′ and q:

q I q′

• Combining all these stetements and using transitivity, we conclude:

q ′′′′ P q ′

Note that this argument can be repeated for any q ′′′′ on the “higher” indifference curve and any q ′ on the
“lower” indifference curve. We conclude: any bundle on a “higher” indifference curve is preferred over any
bundle on a “lower” indifference curve.
There are a couple of other things I want to emphasize about indifference curves:
• Since the preference relation of a consumer is assumed to be monotonic, the indifference curves must
be downward sloping. Suppose, for a contradiction, that a part of an indifference is upward-sloping.

5
 q2

q0
q 00

q q 0000 “higher”
indifference curve

q 000
“lower”
indifference curve

q1

Figure 3: Multiple indifference curves. Any bundle on a “higher” indifference curve is preferred over any
bundle on a “lower” indifference curve.

Then, there are two bundles q and q ′ on the same indifference curve such that q1′ > q1 and q2′ > q2 .
By monotonicity, we must have q ′ P q. But then, q and q ′ cannot be on the same indifference curve! A
contradiction.
• As long as transitivity and monotonicity are satisfied, indifference curves cannot cross. I am leaving
this as an exercise for you to show.

2.2.2 Marginal Rate of Substitution

We will now define a very important object based on the indifference curves. It will provide the very crucial
information on how much the consumer “values” good 1 over good 2 at a certain bundle.
The marginal rate of substitution of good 2 for good 1 at the bundle q = (q1 , q2 ), denoted MRS2,1 (q),
is the rate at which good 2 must substitute for a “small” decrease in the consumption of good 1 in order to
keep the consumer indifferent to the initial bundle q. More formally (see Figure 4):
∆q2
MRS2,1 (q1 , q2 ) = lim = |slope of ind. curve at q| . (1)
∆q1 →0+ ∆q1
(q1 −∆q1 ,q2 +∆q2 )I(q1 ,q2 )

The marginal rate of substitution of good 1 for good 2 at the bundle q, denote MRS1,2 (q), is similarly defined:
∆q1 1
MRS1,2 (q1 , q2 ) = lim = . (2)
∆q2 →0+ ∆q2 |slope of ind. curve at q|
(q1 +∆q1 ,q2 −∆q2 )I(q1 ,q2 )

MRS2,1 (q) is a measure of how much the consumer “values” good 1 in terms of good 2 when he is endowed
with the bundle q. Assume that the consumer is endowed with the bundle q = (q1 , q2 ). If we ask the
consumer to give up a “small” amount of good 1, say ∆q1 units, the consumer would require “approximately”
MRS2,1 (q)∆q1 units of good 2, to compensate the reduction in the quantity of good 1 in order to be indifferent
between the initial bundle and the bundle after the exchange. Similarly, if we asked to consumer to give
up a “small” amount of good 2, say ∆q2 units, the consumer would require “approximately” MRS1,2 (q)∆q2
units of good 1 to compensate the reduction in the quantity of good 2.
Generally speaking, there are four ways to interpret MRS2,1 (q).

6
 q2 q2

M
∆q2 R
S
2,
(q1, q2) 1(
q) (q1, q2)
∆q1 =
−
sl
op
e

q1 q1

Figure 4: Marginal rate of substitution of good 2 for good 1: Taking the limit of ∆q2 /∆q1 as ∆q1 goes to
zero in the figure on left we obtain the figure on the right.

1. (Mathematical.) It is the limit of a ratio of two differences: see equation (1).

2. (Verbal.) It is a measure of how many units of good 2 the consumer must be given, so that she is left
indifferent to a decrease in good 1.
3. (Geometrical.) It is the (absolute value of) the slope of the indifference curve: the steeper the
indifference curve is, the higher MRS2,1 (q) is.
4. (Economical.) It is a measure of the value of good 1 in terms of good 2: the more valuable good 1
is, the higher MRS2,1 (q) is.
Let me now define two more properties on the preference, which are common features of many preferences
in real life.
• Generally, when a consumer has more of a good (say, good 1), and less of another good (say, good 2),
then good 1 becomes less “valuable” for the consumer relative to good 2. We formalize this idea as
follows:
Let q and q ′ be any two bundles that the consumer is indifferent between (i.e., they are on the same
indifference curve). We require preference relations to be such that, if q1 > q1′ , then MRS2,1 (q) <
MRS2,1 (q ′ ). Also, if q2 > q2′ , then MRS1,2 (q) < MRS1,2 (q ′ ).
Preference relations that satisfy this condition are said to have (strictly) diminishing marginal
rate of substitution. If a preference relation satisfies the diminishing marginal rate of substitution
assumption, the relative value of good 1 in terms of good 2 will be lower as the consumer has more of
good 1 and less of good 2. As a result, the indifference curve will get flatter as q1 increases along the
same indifference curve. This means: the indifference curves will be bowed toward the origin (Figure
5).
• A preference relation is said to be smooth if the indifference curves do not have any kinks. The first
two indifference curves displayed in Figure 6 are examples of smooth preference relations and the next
two are examples of preference relations that are not smooth. If a preference relation is smooth then
for any bundle q with positive component (i.e., q1 > 0 and q2 > 0) we have
1
MRS1,2 (q) = . (3)
MRS2,1 (q)

3 Optimal Bundle
Okay, now that we have a grasp of the consumer’s constraints and preferences, it is time to characterize her
choice. The following is a formal definition of the consumer’s “favorite bundle among the feasible ones”.

7
 q2 q2

q
q

q′
′
q
q1 q1

Figure 5: The graph one the left displays a indifference curve on which the diminishing marginal rate of
substitution assumption holds. The graph on the right displays an indifference curve on which the marginal
rate of substitution of good 2 for good 1 increases as we move in the increasing q1 direction along the
indifference curve. Hence the diminishing marginal rate of substitution assumption does not hold for the
indifference curve on the right.

Definition 2. Given the prices p1 , p2 , . . . , pn and income I, a bundle q ∗ = (q1∗ , q ∗ , . . . , qn∗ ) is an optimal
bundle if and only if
Pn
• ∗ ∗ ∗ ∗ ∗
i=1 pi qi = p1 q1 + p2 q2 + · · · + pn qn ≤ I (q is feasible), and
Pn
• for any bundle q = (q1 , q2 , . . . , qn ), if i=1 pi qi ≤ I (i.e., q is feasible), then q ∗ R q (i.e., q ∗ is at least
as good as any feasible bundle).
Thus, a bundle is optimal if and only if it is feasible and is at least as good as (for the consumer) any feasible
bundle. Alternatively, a bundle q ∗ is optimal if and only if any bundle that is preferred to q ∗ is not feasible.
We will now find the optimal bundle q ∗ when there are two goods (the argument is generalizable to more
than two goods). A starting point to develop the intuition is as follows. The optimality conditions imply
that the consumer needs to find the highest indifference curve, given her budget constraint. Can
you try to draw a budget set, the indifference curves, and find the optimal bundle?
Now, let’s get more formal. Please note that all the statements below assume that the preferences are “well-
defined” (i.e., they satisfy completeness, transitivity and monotonicity). In what follows, I will posit some
claims – which are “Claims” in the mathematical sense, so they are correct statements under the assumptions
we made. They are not “claims” in the colloquial sense. They have proofs. I am relegating the proofs to
the Appendix to make this document more readable. Check them out if you are interested.
Assume that there are two goods and q ∗ = (q1∗ , q2∗ ) ∈ R2+ is an optimal bundle. Our first claim is that the
the optimal bundle must be on the budget line.
Claim 1. If q ∗ is optimal, then p1 q1∗ + p2 q2∗ = I.
Informally, Claim 1 means that the consumer must exhaust her budget under the optimal bundle. This is
intuitively due to monotonocity: more is always better than less, and there is no reason to keep the money
in the pocket, so you better just spend the money.
Our second claim is a subtle one that relates the marginal rate of substitution to the price ratio.
Claim 2. If q ∗ is optimal and q1∗ > 0, then
p1
MRS2,1 (q ∗ ) ≥ . (4)
p2

Informally, Claim 2 says the following: if the consumer is buying good 1 in a strictly positive quantity, then
it must be the case that she likes good 1 enough. Otherwise, buying good 1 would not be optimal.

8
 q2 q2

q1 q1
q2 q2

q1 q1

Figure 6: The indifference curves at the top are smooth and the indifference curves at the bottom have kinks.

9
The consumer could also consider consuming a “little” less of good 2 (provided that q2∗ > 0). Arguments
similar to the above would yield the following claim.
Claim 3. If q ∗ is optimal and q2∗ > 0, then
p2
MRS1,2 (q ∗ ) ≥ . (5)
p1

Informally, Claim 3 says: if the consumer is buying good 2 in a strictly positive quantity, then it must be
the case that she likes good 2 enough. Otherwise, buying good 2 would not be optimal.
Now, it is time to combine everything we know and have the “big reveal” of consumer theory. That would
be the theorem below.
Theorem 1. Given the prices p1 , p2 and income I, if q ∗ is an optimal bundle, then p1 q1∗ + p2 q2∗ = I and
• if q1∗ > 0, then
p1
MRS2,1 (q ∗ ) ≥ ,
p2

• if q2∗ > 0, then

p2
MRS1,2 (q ∗ ) ≥ ,
p1

• if q1∗ > 0, q2∗ > 0, and preference is smooth, then

p1
MRS2,1 (q ∗ ) = .
p2

Figure 7 illustrates the optimal bundle when q1∗ > 0 and q2∗ > 0. Intuitively, it is the point where the
indifference curve passing through q ∗ barely touches the budget line, i.e. it is tangent to the budget line.
q2

I
p2

q∗

I
q1
p1

p1
Figure 7: The optimal bundle q ∗ is on the budget line and satisfies MRS2,1 (q ∗ ) = p2 .

Take a moment to appreciate the beauty of this result! In the optimal bundle, the marginal rate of substi-
tution is exactly equal to the price ratio. That is, suppose you go and ask the consumer in a supermarket:
“I see your shopping cart, which contains your optimal bundle. Let me ask you a hypothetical
question. At the optimal bundle, how much do you value good 1 in terms of good 2?”
And suppose her answer is:
“Five. You need to give me five units of good 2 for me to give up one unit of good 1.”

10
And then you go ask the cashier in the supermarket:
“How expensive is good 1 relative to good 2?”
Almost magically, her answer is:
“Five. You can give up five units of good 2 and buy one unit of good 1 instead.”
Isn’t this amazing? What is more amazing is that this applies to every single consumer in the supermarket,
regardless of their preferences. Another customer may be a fan of good 1, i.e. her MRS2,1 may be higher
for every bundle. Fine, she keeps buying more of good 1 and less of good 2, until the marginal rate of
substitution decreases and she finds the bundle where the marginal rate of substitution equals the price
ratio.
In the appendix, I provide some examples of “famous” preferences and discuss the properties of the optimal
bundle under those preferences. You may want to take a look at them before you solve some exercises.

4 Changes in Parameters
Now that we know how the consumer chooses her optimal bundle, we now have the machinery to study how
the optimal bundle changes with the parameters of the model (i.e., income of the consumer and prices of
goods.) This is the fun stuff!

4.1 What If the Income Changes?

Let’s start with an simple case. Consider an increase in a consumer’s income (i.e., I goes up.) What happens?
Mathematically, the optimization problem changes because the constraint set changes. But that’s fine – we
did our analysis using a generic set of parameters, so the analysis still applies. In general, the optimal bundle
q ∗ = (q1∗ , q2∗ ) satisfies

p1 q1∗ + p2 q2∗ = I

Moreover, if q1∗ > 0, q2∗ > 0, and preference is smooth, then

p1
MRS2,1 (q ∗ ) =
p2
So far so good. Just solve this problem with a higher I. Geometrically, it corresponds to shifting the budget
line higher, finding a new indifference curve tangent to it, and marking the point of tangency as the optimal
bundle.
Let’s do this graphically. Notation:
• Fix the prices at p1 and p2 .
• Initial income: I i .
• Optimal bundle under initial income: q i = (q1i , q2i ).
• Final income: I f .
• Optimal bundle under final income: q f = (q1f , q2f ).
If I f > I i , it may look like Figure 8. The red line is the budget line under I i . The dark red line is the budget
line under I f . Note that the two budget lines are parallel, because their slopes are the same: they are − pp12 ,
which we keep fixed for this exercise. The dark red line is higher than the red line, because I f > I i .
The first thing that you should realize is that the consumer is at least as happy as before when she consumes
q f rather than q i . There are two ways in which you can verify this.

11
 q2

qf

qi

p1
q1
+
p2
p1

q2
q1
+

=
p2

I
f
q2
=
I i

q1
q1i q1f

Figure 8: Optimal bundles under I i and I f .

1. The consumer is richer under income I f compared to income I i . This is because I f > I i , but you
can verify this by looking at Figure 8. The budget set under I f is larger than the budget set under
I i . This means that any feasible bundle under I i is also feasible under I f . Therefore, any bundle that
the consumer can afford initially, she can also afford now. This means that the consumer cannot be
worse off! In the worst case, she can consume the same bundle, q i . This implies that q f must be at
least as good as q i , i.e.

qf R qi

2. Just eyeballing Figure 8, you can see that q f is on a higher indifference curve than q i . This is not
surprising: because the consumer is richer, she cannot find her in a lower indifference curve. A higher
indifference curve means that

qf P qi

You may be tempted to say “But isn’t there a third way in which we can verify q f P q i ? Monotonicity?”
My answer is: yes for Figure 8, but not in general. Because one may have: q1f < q1i , but q2f > q2i . In such a
case, monotonicity would not imply a preference between q i and q f . For instance, you may have a case like
Figure 9. You can verify, using bullet points 1 and 2 above, that consumer is at least as happy as before
when she consumes q f rather than q i . It is not due to monotonicity, though!
This begs the question: what is the exact difference between Figure 8 and Figure 9? Here is the answer.
• If the indifference curves are as in Figure 8, the consumer consumes more of good 1 when she has
higher income. We call goods like these normal goods.
Definition 3. Good i is a normal good if the consumer’s consumption of good i increases with the
consumer’s income.
Examples of normal goods: goods that you consume more as you get richer. Cars, iPhones, sweaters,
herbal teas, dishwashers...
• If the indifference curves are as in Figure 9, the consumer consumes less of good 1 when she has higher
income. We call goods like these inferior goods.

12
 q2

qf

qi

q1
q1f q1i

Figure 9: Optimal bundles under I i and I f , when good 1 is an inferior good.

Definition 4. Good i is an inferior good if the consumer’s consumption of good i decreases with the
consumer’s income.
Examples of inferior goods: goods that you consume less as you get richer. Public transportation, rice,
bulgur, instant noodle, instant coffee...
This classification of goods into two categories will be useful later.
You may have two questions at this point.
1. What if I f < I i ? Just switch the labels of I i and I f . The budget line shifts inwards, and the consumer
becomes worse off. If good 1 is a normal good, the consumer’s consumption of good 1 decreases as the
income decreases. If good 1 is an inferior good, the consumer’s consumption of good 1 increases as the
income decreases.
2. What if good 2 is an inferior good? Once again, just switch the labels of goods. If good 2 is a normal
good, the consumer’s consumption of good 2 increases as the income increases. If good 2 is an inferior
good, the consumer’s consumption of good 2 decreases as the income increases.
Below, I summarize what we have discussed so far. The relationship between the consumption of good i
under optimal bundle (qi∗ ) and income I is as follows.

as I ↑... as I ↓...
if i is a normal good, qi∗ ... ↑ ↓
if i is an inferior good, qi∗ ... ↓ ↑

4.2 What If the Price of a Good Changes?

Now, let’s move on to a slightly more complicated case. Consider an increase in the price of good 1 (i.e., p1
goes up.) Notation:
• Fix the income at I and the price of good 2 at p2 .
• Initial price of good 1: pi1 .
• Optimal bundle under initial price of good 1: q i = (q1i , q2i ).

13
 • Final price of good 1: pf1 .
• Optimal bundle under final price of good 2: q f = (q1f , q2f ).
We can conduct a graphical analysis. If pf1 > pi1 , it may look like Figure 10. The red line is the budget line
under pi1 . The dark red line is the budget line under pf1 . Note that the two budget lines are not parallel.
pi pf
The slope of the red line is − p12 , and the slope of the dark red line is − p12 . The budget set under the final
price is smaller than the budget set under the initial price, because pf1 > pi1 .
q2

I
p2

qf
qi
p 1q 1
f

p1
i q1
+ p2

+
p2
q2 =

q2
=
I
I

q1
q1f I q1i I
p1
f pi1

Figure 10: Optimal bundles under prices pi1 and pf1 .

Now, my claim is that the consumer is at most as happy as before when she consumes q f rather than q i .
There are two ways in which you can verify this.
1. The consumer is effectively poorer under price pf1 compared to price pi1 . You can verify this by looking
at Figure 10. The budget set under pf1 is smaller than the budget set under pi1 . This means that some
feasible bundles under pi1 are not feasible under pf1 any more. The purchasing power of the consumer
has decreased, even though she has the same income as before!
2. Just eyeballing Figure 8, you can see that q f is on a lower indifference curve than q i . This is because
the consumer is effectively poorer.
What I am trying to say is: there is an income effect hidden in this graph. In Figure 10, the consumer
reduces her consumption of good 1 from qii to qif due to two reasons.
1. Due to the income effect, the consumer is poorer. If good 1 is a normal good, the consumer reduces
her consumption of good 1.
2. The relative price of good 1 in terms of good 2 is higher! Even if the consumer was not effectively
poorer, she would choose to consume less of good 1 and more of good 2. Why?
Mathematically: Good 2 is now relatively cheaper, so that the consumer can reduce her consumption
of good 1 a little bit and consume a lot of good 2 instead. Recall that the optimal bundle requires
marginal rate of substitution of good 2 for good 1 to be equal to the price ratio. If price ratio is higher,
the marginal rate of substitution is higher. But if the preferences satisfy diminishing marginal rate of
substitution, this is achieved only when the quantity of good 1 is lower and the quantity of good 2 is
higher.

14
 Economically: The trade-off between good 1 and good 2 has changed. Now, in order to consume the
same amount of good 1, the consumer needs to give up more of good 2. That is, the cost of good 1 in
terms of good 2 is higher. Because of this, the consumer is less willing to consume good 1.
In any case, the consumer would substitute some of good 1 with good 2. This would happen even if the
consumer was not effectively poorer. The consumer just finds it optimal to reduce her consumption of
good 1 and increase her consumption of good 2. This is called the substitution effect.
So, in Figure 10, q1f < q1i due to two effects. We want to decompose these two effects: how much is the
reduction in quantity of good 1 due to the consumer being poorer, and how much of it is due to good 1 being
more expensive relative to good 2?
Recall what I said just above: the substitution effect would work towards the reduction in the quantity of
good 1 “even if the consumer was not effectively poorer”. This is the key: how can we think of a consumer who
is not effectively poorer when the price of good 1 changes? The idea is: we will, hypothetically, compensate
the consumer for the price change. That is, we will imagine we increase the consumer’s income up to the
point where, under the new prices, she is exactly as happy as she was before under the old prices.
More formally, we will find a level of income I c such that the following holds. Suppose, under the prices pf1
and p2 , if the consumer’s income was I c , her optimal bundle would be q c = (q1c , q2c ). We want this optimal
bundle to satisfy:

qc I qi

This construction makes sure that the consumer is exactly as happy as before, even though she is consuming
a different bundle. After all, she is indifferent! This means that she is compensated for the increase in the
price of good 1.
We will call I c the compensated income, and q c the compensated demand. The naming choice should
be obvious by now.
Graphically, what we are doing is shifting the dark red line in Figure 10 until it is tangent to the indifference
curve that contains q i . The tangency point is q c .
q2
p 1q 1
f
+ p2
q2 =
I
c

qc

qf
qi
p 1q 1
f

p1
i q1
+ p2

+
p2
q2 =

q2
=
I
I

q1
q1f q1c q1i

Figure 11: Compensated demand for good 1 (q1c ) and compensated income I c .

In Figure 11, the orange line is the budget line under compensated income I c . If the consumer’s income was
I c instead of I, she would consume q c and be exactly as happy as if she was consuming q i . Therefore,

15
 • The move from q i to q c is due to the change in relative prices. It isolates the consumer’s unhappiness
due to being effectively poorer! She is as happy as before, she is just finding it optimal to consume less
of good 1 and more of good 2 because good 1 is relatively more expensive.
• The move from q c to q f is due to the consumer being poorer. There is no effect of relative price, the
consumer just changes her consumption because she is poorer.
Now, the move q c → q f should be familiar to you: this is the income effect. The comparison of q1c and q1f
is the same as before. If good 1 is a normal good, q1f < q1c . If good 1 is an inferior good, q1f > q1c .
But what about the relationship between q1i and q1c ? That is, what is the direction of substitution effect?
My claim is that, as long as diminishing marginal rate of substitution is satisfied, we must have qic < q1i .
Why?
• Intuitively, the move from q1i captures the effect of good 1 being relatively more expensive in terms
of good 2. When something is more expensive, you consume less of it!
• Mathematically, q i in Figure 11 satisfies:

pi1
M RS2,1 (q i ) =
p2
and, by construction, q c satisfies:

pf1
M RS2,1 (q c ) =
p2
pf1 pi1
But since pf1 > pi1 , p2 > p2 . Therefore,

M RS2,1 (q c ) > M RS2,1 (q i )

But recall that q c and q i are on the same indifference curve by construction! Since the preferences
satisfy diminishing MRS, M RS2,1 (q c ) > M RS2,1 (q i ) is satisfied only when q c is to the northwest of
q i . Then, we must have q1c < q1i .
This is what I am saying: as long as the diminishing marginal rate of substitution is satisfied, for any good
i:

as pi ↑... as pi ↓...
the substitution effect is such that qi∗ ... ↓ ↑

But recall that the total effect is a combination of substitution effect and income effect. For a normal good,
let’s put them together.

as p1 ↑...
substitution effect income effect (I ↓) total effect
if i is a normal good, qi∗ ... ↓ ↓ ↓

Both effects work in the same direction! The total effect is a decrease in the quantity of good 1 consumed.
Graphically, this looks like Figure 11. In Figure 12 below, I demonstrate the two effects. Recall that q1i → q1c
is the substitution effect, and q1c → q1f is the income effect. As long as diminishing MRS is satisfied, q1c < q1i .
As long as good 1 is a normal good, q1f < q1c .

16
 q2
I
p2

qc

qf

qi

Inc. Eff. Subs. Eff.

Total Change
q1
q1f q1c I c
q1ipI ′ I
p′1 1
p1

Figure 12: Effect of an increase in the price of good 1, when good 1 is a normal good.

What is good 1 is still a normal good, but it price decreases? You can just imitate the same analysis. All
the effects will be reversed.

as p1 ↓...
substitution effect income effect (I ↑) total effect
if i is a normal good... qi∗ ↑ qi∗ ↑ qi∗ ↑

See Figure 13 below.

17
 q2
I
p2

Ic
p2
qi

qf

qc

Subs. Eff. Inc. Eff.

Total Change
q1
q1i q1c I Ic
f I
p1 p′1 q1 p′1

Figure 13: Effect of a decrease in the price of good 1, when good 1 is a normal good.

The next question is: what if good 1 is an inferior good and the price of good 1 increases? The substitution
effect would still work in the direction of reducing the consumption of good 1. But now, income effect pulls
in the opposite direction.

as p1 ↑...
substitution effect income effect (I ↓) total effect
if i is an inferior good, qi∗ ... ↓ ↑ ?

Hm, this looks like a tricky case. If the income effect dominates, the consumer consumes more of good 1
when it is more expensive! What is happening? The consumer is so poor as a result of the price change that
she moves away from higher quality consumption options and starts consuming good 1 even more.
We economists call such good Giffen goods, named after Robert Giffen. In a letter written to his friend
Alfred Marshall, Giffen suggested the following phenomenon: in the late 19th century, as the price of bread
increased, very poor individuals in Britain consumed more bread! Here is a quote from Wikipedia:
As Mr. Giffen has pointed out, a rise in the price of bread makes so large a drain on the resources
of the poorer labouring families [...] that they are forced to curtail their consumption of meat
and the more expensive farinaceous foods: and, bread being still the cheapest food which they
can get and will take, they consume more, and not less of it.
-Alfred Marshall, 1895
Formally,
Definition 5. Good i is a Giffen good if the consumer’s consumption of good i increases as the price of
good i increases.
Note that for a good to be a Giffen good, it has to be in inferior good: the income effect should pull towards
an increase in the quantity consumed. But being an inferior good is not enough in itself! The good has to
be so inferior that the income effect must dominate the substitution effect! Graphically, it looks like Figure
14.
I would claim that having a Giffen good is a mathematical possibility, but economically it is so unlikely that

18
 q2
I
p2

qc

qi

qf
Inc. Eff.
Subs. Eff.

Total Change q1
q1c q1i q1f I Ic
I
p′1 p′1 p1

Figure 14: Effect of an increase in the price of good 1, when good 1 is a Giffen good.

we can just assume it away. A good such that as it becomes more expensive, you buy more of it! I don’t
find this possibility very compelling. Giffen’s observation about the bread in late 19th century Britain is
controversial: we are not sure it empirically holds. Some claim that potatoes during the great Irish famine
may be considered a Giffen good. Well, maybe, but even if that’s true, that is a very particular time and
location in history. There is a 2008 paper your textbook discusses, which I will post to Moodle. It argues
that in very poor parts of China, rice is a Giffen good. This paper is basically the only empirical evidence
we know about the existence of a Giffen good. But in virtually any economic scenario we consider, the
likelihood of having a Giffen good is so small that we can just discard that possibility. From now on, we will
assume that a good is not a Giffen good. It may still be an inferior good, but even then we will assume that
the income effect does not dominate the substitution effect. Those goods are sometimes called ordinary
goods.
Definition 6. Good i is an ordinary good if the consumer’s consumption of good i decreases as the price
of good i increases.
From now on, let’s agree that a good is not a Giffen good.

19
 q2
I
p2

qc

qi

qf

Subs. Eff.
Inc. Eff.
Total Change q1
q1c q1f I q1i Ic
I
p′1 p′1 p1

Figure 15: Effect of an increase in the price of good 1, when good 1 is an inferior but not a Giffen good.

Appendix
A Proofs
Proof of Claim 1. Suppose, towards a contradiction, that p1 q1∗ + p2 q2∗ ̸= I. There are two possibilities.
• If p1 q1∗ + p2 q2∗ > I, q ∗ would not be feasible. This would contradict feasibility of q ∗ .
• If p1 q1∗ + p2 q2∗ < I, the consumer can afford another bundle q ′ = (q1∗ + ∆1 , q2∗ + ∆2 ), i.e. q ′ is feasible.
Because the preference relation is monotonic, this is preferred to q ∗ , i.e. q ′ P q ∗ . This contradicts the
optimality of q ∗ .
Since both cases lead to a contradiction, the proof follows.

Proof of Claim 2. Assume that q1∗ > 0. Let ∆q1 be a “small” positive quantity such that q1∗ − ∆q1 ≥ 0.
(Since q1∗ is positive, there is such a positive quantity).
The consumer is considering the bundle q ∗ . If she consumes ∆q1 units less of good 1 (i.e., consumes q1∗ − ∆q1
units of good 1 rather than q1∗ units of it), then the consumer would require (approximately) MRS2,1 (q ∗ )∆q1
units of good 2 to substitute for good 1 (so that the she is indifferent between the final bundle and the initial
bundle q ∗ ). But if the consumer buys ∆q1 units less of good 1, she would have p1 ∆q1 TL to spend on good 2.
With this money she can buy (p1 ∆q1 )/p2 units of good 2. If
p1 ∆q1
MRS2,1 (q ∗ )∆q1 < , (6)
p2
then the consumer would become better off by consuming ∆q1 units less of good 1 and (p1 ∆q1 )/p2 units
more of good 2. That is, if (6) holds, then the bundle
(q1∗ − ∆q1 , q ∗ + (p1 /p2 )∆q1 )
is feasible and is preferred to the bundle q ∗ . But this contradicts with q ∗ being optimal. Thus, if q ∗ is
optimal, then (6) can not be true, which means that
p1 ∆q1
MRS2,1 (q ∗ )∆q1 ≥
p2

20
must be true. Since ∆q1∗ is positive, dividing both sides of the above inequality with q1∗ we obtain:
p1
MRS2,1 (q ∗ ) ≥ .
p2

Proof of Claim 3 is very similar to that of Claim 2, so I am leaving it as an exercise.

Proof of Theorem 1. The proof of first two bullet points follow from Claims 2 and 3. For the last bullet
point: If preference is smooth and q ∗ is an optimal bundle with q1∗ > 0 and q2∗ > 0, then Claim 2, (3), and
Claim 3 imply:
p1 1 p1
≥ = MRS2,1 (q ∗ ) ≥
p2 MRS1,2 (q ∗ ) p2
Which in turn implies
p1
MRS2,1 (q ∗ ) = .
p2

B Optimal Bundle for Some Famous Preferences

(This appendix is meant to be supplementary. It will hopefully serve as a guideline for future exercises.
Please take some personal time to go through these examples on your own.)
The preferences we consider throughout this appendix satisfy completeness, transitivity and monotonicity.
We will keep assuming that there are two goods for the sake of visualization, but once again the ideas extend.
We will keep the budget constraint the same across examples: a bundle q = (q1 , q2 ) is feasible if and only if
p1 q1 + p2 q2 ≤ I.

B.1 Perfect Substitutes

Suppose the consumer’s preferences are such that: for any q = (q1 , q2 ) and q ′ = (q1′ , q2′ ),
q R q ′ ⇐⇒ aq1 + bq2 ≥ aq1′ + bq2′ (7)

where a > 0 and b > 0.

How do indifference curves look like? Recall that the consumer is indifferent between any two bundles
on an indifference curve. Therefore, if q and q ′ are on the same indifference curve,
q I q ′ ⇐⇒ q R q ′ and q ′ R q (by definition of indifference)
⇐⇒ aq1 + bq2 ≥ aq1′ + bq2′ and aq1′ + bq2′ ≥ aq1 + bq2 (by the preferences in Equation (7))
⇐⇒ aq1 + bq2 = aq1′ + bq2′

What does it mean? Consider the line defined by the equation:

aq1 + bq2 = c (8)

with some c ≥ 0. The consumer is indifferent between any two bundles q and q ′ on this line, because
aq1 + bq2 = c = aq1′ + bq2′

Let’s make sure that this is actually a line. Rearranging Equation (8), we arrive at:
c a
q2 = − q1
b b

21
 c
which is, geometrically, the equation for a line with intercept b and slope − ab .
So the indifference curves in this case are parallel lines, each with slope − ab . A higher c means that
the consumer is on a “higher” indifference curve, meaning that the consumer prefers the bundles on the
indifference curves with higher c to the bundles on the indifference curves with lower c. You can verify this
using two alternative methods.
1. Take two indifference curves

aq1 + bq2 = c
aq1 + bq2 = c′

with c′ > c. Just drawing these indifference curves, you will see that the second indifference curve is
higher than the first one (it is further away from the origin.)
Take a bundle q = (q1 , q2 ) on the first indifference curve, and another bundle q ′ = (q1′ , q2′ ) on the second
indifference curve. By the equations defining the indifference curves, they following equalities must
hold:

aq1 + bq2 = c
aq1′ + bq2′ = c′

But since c′ > c, aq1′ + bq2′ > aq1 + bq2 . Then, by the preferences in Equation (7), q ′ P q.
2. Just pick two bundles q = (q1 , q2 ) and q ′ = (q1′ , q2′ ) where q1′ > q1 and q2′ > q2 . Draw the indifference
curves that pass through q and q ′ , and you will see that the one that passes through q ′ is further away
from the origin. By monotonicity,

q′ P q

By the definition of an indifference curve, the consumer is indifferent between q and any bundle q ′′ on
the indifference curve passing through q.

q ′′ I q

Similarly, the consumer is indifferent between q ′ and any bundle q ′′′ on the indifference curve passing
through q ′ .

q ′′′ I q ′

By transitivity,

q ′′′ I q ′ P q I q ′′ =⇒ q ′′′ P q ′′

Fine, but what do they mean? The interesting thing about lines is that their slopes are constant. Since
the (absolute value of the) slope of the indifference curve is the marginal rate of substitution, it means that
the marginal rate of substitution is constant.
a
M RS2,1 (q) = for all q
b

Recall that M RS2,1 (q) is a measure of how much the consumer values good 1 in terms of good 2 when she is
endowed with bundle q. When M RS2,1 (q) is constant, this means that the relative value of good 1 does not
depend on the bundle q. No matter how many units of good 1 and good 2 the consumer considers, the relative
value is the same. You can always take away b units of good 1 from the consumer, compensate
the consumer by giving a extra units of good 2, and leave the consumer indifferent. That is, no
matter what the consumer is endowed with, a units of good 2 can always perfectly substitute b units of
good 1. That’s why good 1 and good 2 are perfect substitutes.

22
 q2
c0

slo
pe
q 00 q0

=−
q2

a
b
q

q 000

q1 q1

Figure 16: Indifference curves for preferences given in Equation (7).

Examples? We typically consider the goods that are very similar in quality to be perfect substitutes. Take
Coca Cola and Pepsi, for instance. You may like Coca Cola more than Pepsi, which is fine. In that the, the
marginal rate of substitution of Pepsi for Coca Cola will be higher than one. What matters is that if you
are willing to substitute one Coca Cola for one Pepsi when you have 10 Pepsis and 0 Coca Colas, then you
should be willing to substitute one Coca Cola for one Pepsi when you have 9 Pepsis and 1 Coca Cola. So
and so on.

Is the diminishing marginal rate of substitution satisfied? No. The marginal rate of substitution
is constant.

Are the preferences smooth? Yes. The indifference curves do not have any kinks. Therefore, M RS1,2 (q) =
1 1 b
M RS1,2 (q) = a/b = a for all q.

a p1
What is the optimal bundle? Depends on b (marginal rate of substitution) and p2 (marginal rate of
transformation.)
• To begin, suppose ab < pp12 and consider the optimal bundle q ∗ = (q1∗ , q2∗ ). That is, the indifference
curves are flatter than the budget line. I claim that in this case, we must have q1∗ = 0. Why?
Suppose not, i.e., suppose q1∗ > 0. But then, by Theorem 1, we must have M RS2,1 (q ∗ ) ≥ pp12 . But recall
that M RS2,1 (q ∗ ) = ab < pp21 . This is a contradiction. Therefore, we cannot have q1∗ > 0. We conclude
that q1∗ = 0, and the consumer spends all her income on q ∗ 2. The optimal bundle is q ∗ = (0, pI2 ).
Intuitively, what is going on? ab being low means that the consumer does not value good 1 much.
Indeed, the relative price of good 1 in terms of good 2 is higher than the relative value of good 1 in
terms of good 2. The consumer can always buy b units less of good 1. This will save the consumer bp1 .
With these savings, the consumer can buy an extra bp p1 a bp1
p2 units of good 2. Since p2 > b , p2 > a. Thus,
1

the consumer can buy more than a units of good 2 with her savings. But remember that a units of
good 2 would leave the consumer indifferent! Anything more than a units of good 2 would make the
consumer strictly happier! As a result, the consumer keeps reducing her consumption of good 1 until

23
 she has no good 1 left in her bundle.
Geometrically, the following is going on:
q2

q∗

slo
slope =

pe
=−
a
b
− p2
p1

q1
a p1
Figure 17: Optimal bundle when b < p2 . The blue lines are indifference curves and the red line is the budget
line.

It’s just the consumer finding the “highest” indifference curve subject to the budget constraint.
• Next, suppose ab > pp12 , i.e. the indifference curves are steeper than the budget line. Consider
the optimal bundle q ∗ = (q1∗ , q2∗ ). I claim that in this case, we must have q2∗ = 0. Why? Suppose
not, i.e., suppose q2∗ > 0. But then, by Theorem 1, we must have M RS1,2 (q ∗ ) ≥ pp12 . But recall that
M RS1,2 (q ∗ ) = ab < pp21 . This is a contradiction. Therefore, we cannot have q2∗ > 0. We conclude that
q2∗ = 0, and the consumer spends all her income on q1∗ . The optimal bundle is q ∗ = ( pI1 , 0).
• Finally, consider the case ab = pp12 . The indifference curves are parallel to the budget line! In this case,
there are many optimal bundles. Indeed, any bundle on the budget line is optimal.
This is a somewhat knife-edge case (a consumer whose relative value exactly equals the relative price),
but not impossible to find. For instance, suppose a = b and p1 = p2 . This means the consumer is
totally indifferent between the goods (take away good 1, give good 2 in equal amounts, doesn’t matter),
and also the price are equal. This corresponds to cases where the brand of the item does not matter,
at all. I am thinking of something like bleach. Does the brand of the bleach matter at all? For many
people, no. When I need to buy bleach, I would just go ahead and buy the cheapest one. If the two
brands in the supermarket have the same price, I could buy either of them.

B.2 Quasi-linear Preferences

Suppose the consumer’s preferences are such that: for any q = (q1 , q2 ) and q ′ = (q1′ , q2′ ),

q R q ′ ⇐⇒ v(q1 ) + q2 ≥ v(q1′ ) + q2′ (9)

where v(x) is an increasing and concave function. That is, its first derivative is positive and decreasing (it
second derivative is negative). Think of v(x) = log x or v(x) = xα for some α ∈ (0, 1).

24
 q2

q∗ q1

a p1
Figure 18: Optimal bundle when b > p2 . The blue lines are indifference curves and the red line is the budget
line.

How do indifference curves look like? If q and q ′ are on the same indifference curve,

q I q ′ ⇐⇒ q R q ′ and q ′ R q (by definition of indifference)

⇐⇒ v(q1 ) + q2 ≥ v(q1′ ) + q2′ and v(q1′ ) + q2′ ≥ v(q1 ) + q2 (by the preferences in Equation (9))
⇐⇒ v(q1 ) + q2 = v(q1′ ) + q2′

What does it mean? Consider the curve defined by the equation:

v(q1 ) + q2 = c (10)

This is a typical indifference curve for quasi-linear preferences, where higher values of c correspond to “higher”
indifference curves. You can rearrange this to get:

q2 = c − v(q1 ) (11)

Since v(x) is increasing, this curve is downward-sloping. Since v(x) is concave, this curve is convex. For a
higher c, we shift this curve upwards.

Fine, but what do they mean? As you can guess by its name, quasi-linear preferences are “kind of”
like linear preferences. By “kind of”, we mean preferences are linear with respect to one good (in this case,
good 2) and not linear with respect to the other good (good 1).
The marginal value that the consumer assigns to good 1 is decreasing in the amount of good 1 the consumer
has. It is, however, constant in the amount of good 2 that the consumer has. When the consumer is endowed
with more of good 1, she starts liking it less – this is like a usual good we consider. When the consumer is
endowed with more of good 2, her attitudes towards good 2 does not change – this is like “linear” preferences.
You can see this feature by checking the marginal rate of substitution. Just take the derivative of Equation
(11) and take its absolute value to find the slope of the indifference curve:

M RS2,1 (q) = v ′ (q1 ) for all q = (q1 , q2 )

25
 q2

q∗ q1

a p1
Figure 19: Optimal bundles when b = p2 .

As you see, this depends on q1 but not on q2 . As long as q1 remains the same, you can increase q2 and
M RS2,1 (q) does not change. This means if you compare two indifference curves and keep q1 constant, their
slopes are the same. Therefore, indifference curves are just shifted versions of each other in the y-axis.

Examples? Good 1 in this example is a standard consumption good, like apples. Good 2 in this example
is a good so that the consumer’s feelings towards it does not change no matter how much of it she has. Let
me give a somewhat radical example. Consider good 2 as money. The price of good 2 is p2 = 1. That is,
you can spend 1 TL and buy one unit of good 2, which is again 1 TL. Of course, this is just a representation:
we are not considering a consumer who spends money to buy money. Instead, we are thinking about a
consumer with a certain budget who decides how many apples to buy (q1 ) and how much money to keep in
her pocket (q2 ). The crucial thing is that 1 TL is always 1 TL, no matter how much money you have. So it
is reasonable to assume that consumers’ feelings towards money does not depend on how much money they
have already.1
I just want to point out: our earlier discussions made it look like the consumer has to spend all her income
when she goes shopping. Now you realize that this framework allows for more general outcomes. It is possible
to introduce money saved as another good and conduct the analysis as usual.
Another example: let q1 denote the time spent on studying for the economics exam, and q2 denote the time
spent on other activities (such as watching more episodes of Ask-i Memnu). v(q1 ) is the expected grade on
the economics exam when the student studies for q1 hours. The student has a very standard thing to do
when she does not study (the satisfaction you derive from Ask-i Memnu neither goes up nor goes down as
you watch more of it), so the value of the alternative activities does not change at all.

Is the diminishing marginal rate of substitution satisfied? Yes. Recall that v ′ (q1 ) is decreasing.

Are the preferences smooth? Yes, as long as v(x) is a smooth function (it does not have kinks).
1 This does not have to be universally correct: you can imagine people valuing the extra lira less if they have more money

already. That is, the preferences towards money can also satisfy diminishing marginal value. But especially for cash-constrained
consumers, having quasi-linear preferences with respect to money seems like a reasonable thing to do.

26
 q2

sl o
pe
=
−v
0 (x
)

x q1

Figure 20: Indifference curves for preferences given in Equation (9).

What is the optimal bundle? The bottom line is that: the consumer keeps buying good 1 until the
point where her marginal rate of substitution is low enough, so she does not want to buy it any more. The
question is: what is the marginal rate of substitution if she, hypothetically, has spent all her income on good
1? At this point, she has bought pI1 units of good 1. If this much is enough so that M RS2,1 (q) is lower than
p1
p2 , that is more than enough. She should have stopped buying earlier and spend the remaining amount on
good 2. If M RS2,1 (q) > pp12 at this bundle, she spends all her income on good 1. If she had even more income
she would buy even more of good 1, but she is constrained, so she stops here.
p1
Formally, the optimal bundle depends on v ′ ( pI1 ) (marginal rate of substitution) and p2 (marginal rate of
transformation.)
• If v ′ ( pI1 ) ≤ p1
p2 , the optimal bundle is q ∗ = (q1∗ , q2∗ ) such that
p1 p1
M RS2,1 (q ∗ ) = =⇒ v ′ (q1∗ ) =
p2 p2
I−p1 q1∗
and q2∗ = p2 It’s easier to see graphically; see Figure 21.

• If v ′ ( pI1 ) > p1
p2 , the optimal bundle is q ∗ = ( pI1 , 0). See Figure 22.

B.3 Cobb-Douglas Preferences

The following preferences are “invented” by Charles Cobb and Paul Douglas in the first half of 20th century.2
Suppose the consumer’s preferences are such that: for any q = (q1 , q2 ) and q ′ = (q1′ , q2′ ),
q R q ′ ⇐⇒ (q1 )α (q2 )1−α ≥ (q1′ )α (q2′ )1−α (12)

where α ∈ [0, 1] is a parameter that measures the “weight” that the consumer attaches to good 1 in her
preferences. If α is higher, consumer has a higher inclination towards good 1.
2 Fun fact: Paul Douglas later went on to serve as a senator ın the US for eighteen years! We, as economists, sometimes wish
that he pushed for a legislation that requires every preference to be Cobb-Douglas. :)

27
 q2

sl o
pe
sl o

=
q2∗
pe

−v
0 (q 1
=

∗)
− p2
p1

q1∗ I
q1
p1

p1
Figure 21: Optimal bundle when v ′ ( pI1 ) ≤ p2 .

How do the indifference curves look like? You can just imitate the arguments in the previous examples
to derive the equation for a typical indifference curve:
(q1 )α (q2 )1−α = c (13)
Once again, higher values of c correspond to “higher” indifference curves. You can rearrange this to get:
1 −α
q2 = (c) 1−α (q1 ) 1−α (14)
You can check that this is downward-sloping.

Fine, but what do they mean? Not much in particular. Cobb-Douglas preferences are the standard
preferences used to capture preferences towards two standard consumption goods. As you will see (and as
we discussed in the lecture), these preferences satisfy all the nice properties of a typical preference relation.
As you will also see, the optimal bundle satisfies certain nice properties.
With a little bit of messy algebra (which you don’t need to know by heart), you can verify that:
α q2
M RS2,1 (q) = for all q = (q1 , q2 ) (15)
1 − α q1
So, the marginal rate of substitution depends only on the ratio of q2 and q1 . Note that if q1 decreases and
q2 increases, the marginal rate of substitution increases, i.e. good 1 becomes relatively more valuable to the
consumer. This is the diminishing marginal rate of substitution!
Also note that the marginal rate of substitution is higher when α is higher, i.e., when the “weight” of good
1 is higher. This makes sense: if the weight of good 1 is higher, the consumer values good 1 more, which
translates into a higher M RS2,1 (q).
More importantly, as long as qq21 remains the same, the consumer’s relative valuation of the good is the same.
Suppose the goods are coffee and eggs. When the consumer is endowed with one cup of coffee and three
eggs, she has a relative value attached to coffee versus eggs. If the consumer has Cobb-Douglas preferences,
then she would have the same relative value when she has two cups of coffee and six eggs.

28
 q2

q∗
q1
I
p1

p1
Figure 22: Optimal bundle when v ′ ( pI1 ) > p2 . The consumer spends all her income on good 1.

Examples? Usual consumption goods. Tea versus coffee. Apples versus bananas. White shirts versus blue
shirts. Sujuk versus Halloumi cheese.

Is the diminishing marginal rate of substitution satisfied? Yessss.

Are the preferences smooth? Yessss.

What is the optimal bundle? Take my word for it when I say that in the optimal bundle q ∗ = (q1∗ , q2∗ ),
the consumer has q1∗ > 0 and q2∗ > 0. Why? If q2∗ = 0 and q1∗ > 0, the consumer would have M RS2,1 (q ∗ ) = 0.
This is inconsistent with q1∗ > 0, as we showed in Theorem 1.
Given that q1∗ > 0 and q2∗ > 0, Theorem 1 yields:
p1
M RS2,1 (q ∗ ) =
p2

Use Equation (15) to substitute the left hand-side:

α q2∗ p1
=
1 − α q1∗ p2

Rearrange to get:
q1∗ p1 α
=
q2∗ p2 1−α

What does this mean? q1∗ p1 is the consumer’s total expenditure on good 1. q2∗ p2 is the total expenditure on
good 2. Combine this with the equation q1∗ p1 + q2∗ p2 = I to derive the following result:

29
 q2

q1

Figure 23: Indifference curves for preferences given in Equation (12). Note the interesting feature. If you
draw a ray passing through the origin, this ray intersects each indifference curve once. Along the ray, qq21 is
constant. Therefore, at those intersections, M RS2,1 (q) will be the same.

“In the optimal bundle, the consumer spends α fraction of her income on good 1, and 1 − α
(1−α)I
fraction on good 2. Therefore, q ∗ = ( αI
p1 , p2 ).”
Circling back to what we had before: recall that α is the weight of good 1. If α is higher, the consumer
allocates a larger share of her budget to good 1!

B.4 Perfect Complements

I will not write this one in much detail. For perfect complements, usually a visual inspection suffices.
Suppose the consumer’s preferences are such that: for any q = (q1 , q2 ) and q ′ = (q1′ , q2′ ),

q1 q2 q′ q′
q R q ′ ⇐⇒ min{ , } ≥ min{ 1 , 2 } (16)
a b a b
where a > 0 and b > 0.
A typical indifference curve is defined by the following equation:
q1 q2
min{ , }=c (17)
a b
Meaning? a units of good 1 perfectly complement b units of good 2. If the consumer has a units of good
1 and more than b units of good 2, the extra units of good 2 are useless. Examples? Left and right shoes,
coffee and sugar (if you are drinking coffee only with sugar and if coffee is the only thing you put sugar in).

Is the diminishing marginal rate of substitution satisfied? Not really.

Are the preferences smooth? Nope.

What is the optimal bundle? See Figure 25. Theorem 1 does not apply in this case because the
preferences are not smooth. But a visual inspection suffices.

30
 qsugar
q1 = 3q2

qcof f ee
1 2
Figure 24: An example of indifference curves for preferences given in Equation (16). Here, good 1 is coffee
(quantity is in cups) and good 2 is sugar (quantity is in cubes). The consumer consumes each cup of coffee
with three cubes of sugar (I know – she should reduce her sugar consumption.) Therefore, one cup of coffee
perfectly complements three cubes of sugar. Thus, a = 1 and b = 3.

qsugar
q1 = 3q2
10

6
q∗

qcof f ee
1 2 5
Figure 25: Optimal bundle. Suppose p1 = 5 TL, p2 = 2.5 TL and I = 25 TL. The consumer could buy 5
cups of coffee, but it will be worthless without the sugar. She could buy 10 cubes of sugar (how expensive
is sugar, by the way???), but that would be worthless without the coffee. In the optimal bundle, she buys 2
cups of coffee with 6 sugars, which perfectly complement each other.

31