Consumer Behavior and Budget Constraints
Consumer Behavior and Budget Constraints
Consumer Behavior
1
Introduction
Consider the following situation. You are an economist for Company A that sells
smartphones and tablets. The company is deciding whether they should increase
the price of tablets or smartphones. As a resident economist, you are asked to
analyze which price should the company increase and by how much? How would
you solve such a problem?
Let us further assume that the tablets (think of Apple iPad) are mostly bought
by college students and smartphones are meant for business executive (think of
iPhone 12 phones). If the price of tablets increase how do you think the revenue
of the company would be affected? How about the price of smartphones? What
information would you need to make such comparisons? What would be the optimal
pricing policy in this case?
Consider another situation. Now you are advising the federal government on
whether to impose a tax on gas. However, you do not know how much gas each
individual customer is buying but you have the price of gas and GDP data from past
years. How would you determine how much tax to impose? How would compare a
tax on gas to an increase in income tax that generates the same revenue?
In the following chapters in this section, we would try to answer these questions.
To answer the first question we need information about demand for smartphones
and tablets by the consumer and also we need to know how the demand changes
with price.
To answer these questions we will assume the agent or the consumer here is ra-
tional in the sense he chooses the best possible alternative for himself. For example,
if he wants to buy a smartphone he would not buy a tablet when both goods are
available to him. This implies the consumer will choose alternatives optimally given
what he likes and what is available to him. To understand what is available to the
consumer we need to know how much money(resources) the consumer has and what
is the price of the commodity. We will discuss the solution to these questions in the
first four chapters.
Answering the second question is harder because we have fewer data points
available to us. This would imply we need to build models that can deal with the
real-life scenario where the economist has to design policies with only price and
demand/choice data from previous periods. We will discuss these questions in the
5 chapter.
Finally, in the last chapter, we would discuss the effects of changing prices and
other factors in the economy to the demand.
3
4
Chapter 1
Budget Set
5
6 CHAPTER 1. BUDGET SET
With the strictly positive income m and strictly positive prices p~ we define the
budget constraint or the budget set of the DM as follows:
p 1 x1 + p 2 x 2 ≤ m (BC)
Any commodity bundle that satisfies equation BC is called a feasible bundle. The
boundary of the budget set is called the budget line and is given by,
p 1 x1 + p 2 x2 = m
If a consumer chooses a bundle on the budget line then he uses his entire income
m on the two goods. One nice property of the budget set thus defined is that for a
given set of prices and income it is a convex set, i.e., all convex combination of any
two elements of the set remain in the set. However, if price changes for different
levels of goods it will not remain so. In the next few sections we will explore cases
where the price of one or both goods change exogenously in the market.
We have taken the price and income as given, they are called the parameters of
the model. We are interested to know how the choice of the consumer will change
when the parameters of the model are changed. To do so, we first ask the question,
what happens to the budget set when we change the parameters of the model.
If income m increases, then the slope of the line does not change but both the x1
and x2 intercept would change. If m increases then both pm1 and pm2 increases, because
now the agent can buy more of both goods with the additional income.(Refer fig
1.2)
Figure 1.3: Panel a: shift in budget line after p1 decreases, Panel b: shift in budget
line after p2 decreases
If the price of good 1 changes, then the slope of the line changes and the x1
intercept changes as well. If p1 decreases, then the slope becomes less negative or
the budget line becomes flatter. In terms of x1 intercept, since income is fixed and
p1 has gone down, pm1 increases, i.e., the agent can buy more amount of good 1 when
8 CHAPTER 1. BUDGET SET
he spends his entire income on good 1. Similarly if p2 decreases the budget line will
become steeper and the intercept pm2 will increase. (Refer fig 1.3)
p1 x1 + p2 x2 = m − T.
The diagram below (refer 1.4) illustrates, For value-added tax (also known as ad
valorem tax), the tax rate applies to the value of good, namely it becomes (1 + τ ) p1 .
So in both cases, the slope of the budget line changes. An example of a value-added
tax is sales tax. The budget constraint, in this case, is given by
(1 + τ ) p1 x1 + p2 x2 = m
The diagram below (refer 1.5) illustrates the change in budget set. In the case of
quantity taxes, if the tax rate on good 1 is τ , then the new price would be p1 + τ .
The budget constraint is given by,
(p1 + τ )x1 + p2 x2 = m
1.4. TAX POLICIES 9
Think about how a quantity tax will change the budget set. Beside taxes we can
also consider subsidies as well, where the effective price is reduced, so the signs in
front of the tax rate term would be negative but the analysis would remain the
same.
Also, the government can have rationing policies that restrict the consumer to
buy only up to k units, in which case at x1 = k the budget line becomes verti-
cal.(Refer fig 1.8)
Thought Exercise
Another important government policy is food stamp or food subsidy program which
runs in many countries. Under this program, the government gives f units of food
stamps to people to buy food. The stamps can only be used to buy food and nothing
else and for purchasing food the stamps are as good as money. If x1 denotes the
consumption of food by an agent then he can buy m+f p1
units of x1 but when he
m
spends all his income on good 2 he can only consume p2 units of good 2. Even
when he spends all his income on good 2, he can use his food stamps and get pf1 .
Draw a budget line where m = $200, f = $40 and the prices are p1 = $5 and
p2 = $10 respectively.
For the case where x1 ∈ [0, 6] the slope of the budget line is −2 and if the
consumer buys only good 2 he can afford x2 = 20. At x1 = 6 the consumer has
already spent $120 on good 1. Hence the maximum amount of good 2 he can afford
is x2 = (200 − 120)/10 = 8 and the maximum amount of additional good 1 he can
afford (above the first six units bought at $10 per unit) is 80/10 = 8. The slope
of the budget line for this case is −1, which is flatter than the initial slope. This
generates the following budget line.
The budget line is piece-wise linear since for different values of x1 the slope
changes. As we will see in later chapters this will complicate the optimization
problem for this consumer.
12 CHAPTER 1. BUDGET SET
Chapter 2
The goal of neoclassical consumer theory is to predict choices from a budget set.
For that, we need a method of selection by the DM from the budget set which
would be general across all prices and income. To begin with, we will construct the
theory of choices over simple arbitrary choice set X and then we will extend the
theory to budget sets. We will start with pairwise comparisons for objects in X
and then extend it to global choice over the entire set X. The goal of the theory
is formalize the idea that if a consumer chooses an element x ∈ X then x must be
the subjectively best choice for him.
13
14 CHAPTER 2. PREFERENCES AND CHOICES
Thought Exercise
Show that the following property of a preference order follows from Completeness.
Reflexive: Any bundle x ∈ X is at least as good as itself.
Claim 1: is transitive
Let x y and y z. To prove is transitive we need to show x z, i.e., x % z
and ¬z % x. Since x y, we know x % y and ¬y % x, similarly y % z and ¬z % y.
By transitivity of % x % z. If is not transitive and z % x then by transitivity of
% z % y since x % y. This contradicts our assumption y z. Hence, proved.
Thought Exercise
Show that ∼ is transitive but not complete.
C(X) = {x ∈ X|x % y, ∀y ∈ X}
The chosen set C(X) can contain more than one element and the theory is silent
about how to chose uniquely in that case.
The assumptions on % ensure that C(X) 6= ∅, i.e., C(X) has at least one
element. We will consider examples with two-element and three-element choice set
to illustrate this. Let us consider the choice X = {x, y, }. Given CT there are three
possibilities. First, x % y only, second x % y and y % x and third, y % x only. In
the first case, we have x y so x will be chosen. In the second case x ∼ y so either
x or y can be chosen. In the third case, y x so y is chosen. Thus the assumptions
guarantee choice of at least one element from a two-element choice set.
Now consider a three-element budget set X = {x, y, z}. If we have x % y, y % z,
and x % z then the preference is complete and transitive. From % we get x y,
y z and x z, which implies the theory suggest to choose x. Similarly for the
same X if we have x % y, y % x, x % z, and y % z, we get x ∼ y z, so the theory
is to choose x or y. This preference is also complete and transitive (How?).
Now consider the same three-element choice set X with only y % x, then the
preference is not complete, since we don’t know anything about z. Thus the theory
2.3. INDIFFERENCE CURVE 15
Math Aside
The transitivity assumption implies that two indifference sets can not intersect. We
will prove this by contradiction. Assume an element x ∈ X = RN + belongs to two
difference indifference sets I1 and I2 . Let y ∈ I1 and z ∈ I2 such that either y z
or z y, i.e., I1 6= I2 . Since x ∈ I1 , we have x ∼ y or x % y and y % x. But x ∈ I2
as well so x ∼ z or x % z and z % x. By transitivity y % z and z % y or y ∼ z,
leading to a contraction.
For any point x ∈ Rn+ we can define the weakly preferred set as the set of all
y ∈ RN N
+ such that y % x. A weakly preferred set of any point x ∈ X = R+ consists
of several indifference sets. In particular, the boundary of the weakly preferred set,
i.e., the set of all points y ∼ x is called the indifference curve (IC). Transitivity
implies that two ICs cannot intersect. The figure below (refer figure 2.1) illustrates,
16 CHAPTER 2. PREFERENCES AND CHOICES
The two regularity conditions below ensure that the indifference curve is well-
behaved, i.e., downward sloping and convex (bowed in).
Strict Monotonicity: A preference relation % is strictly monotone if for any
two elements x̃, ỹ ∈ X = RN + , x̃ > ỹ implies x y. Where x̃ > ỹ implies xi ≥ yi
for all 1 ≤ i ≤ N and there exists at least one j such that xj > yj , that is x̃ is
at least as high as ỹ component-wise and strictly higher in at least one dimension.
This assumption says more is better, i,e., all goods are good and not bad.
Strict monotonicity implies if we go to the north-east of a bundle x̃ ∈ X = RN +
we get bundles that are strictly preferred, which implies the IC can not be upward
sloping, or thick or increasing in south-west direction. Let us show in the following
diagrams. For all three cases we can find two bundles x̃ and ỹ such that y1 > x1
and y2 > x2 . Then by strict monotonicity (y1 , y2 ) (x1 , x2 ) so they cannot belong
to the same IC.
Convexity: The preference relation % is convex if the weakly preferred set for
any bundle is convex, that is all convex combination of any two points in the weakly
preferred set belong to the set itself. Consider two elements x and y such that y ∼ x,
then any convex combination with a weight λ ∈ [0, 1], namely λx+(1−λ)y % x ∼ y.
If the convex combination λx + (1 − λ)y x ∼ y then the preference relation is
strictly convex. Convexity implies that the average bundle is more preferred to the
extreme bundles.
2.3. INDIFFERENCE CURVE 17
2.3.1 Examples:
Perfect Substitutes: Suppose you need to choose between two alternatives to go
to the city museum, one is a red bus and another is a blue bus. The two buses
take exactly the same route and are equally comfortable. This is an example of
perfect substitutes since you would be happy to substitute one with the other. The
diagram below shows the shape of the IC for a perfect substitute.
Perfect Complements: The polar opposite case of perfect substitute would
be perfect complements. An example would be left and right pair of shoes. Unless
you have both of them you can not use the pair. So one extra left shoe is as good
as no extra shoes at all. The L-shaped IC for perfect complements reflects this
complementarity.
means this bundle is weakly preferred to all other bundles. The ICs away from the
satiation point are less preferred. Note that, here the IC is not always downward
sloping.
more and more books in Latin would not be preferred by you and hence would be
a neutral commodity. The diagram below illustrates.
Demand
In this chapter, we will graphically explore optimal choice from a standard budget
set. Throughout the chapter we will assume the choice set with two goods only,
namely, X ∈ R2+ . Recall, in the two good case the standard budget set is given by
p1 x1 + p2 x2 = m.
We will assume the two regularity conditions, namely, strict monotonicity and con-
vexity, defined in the last chapter. No other assumptions are made on the particular
form of the indifference curve, which implies any such curve will be consistent with
the standard choice theory.
Given the preference of the agent and the budget constraint he is facing, we want
to know what the agent would actually choose from his budget set. The implicit
assumption in the standard choice theory is that the agent is rational, which in this
context means that he would choose his most preferred alternative that is feasible
in his budget set. We call this the choice problem for the consumer. In this chapter
we will solve the choice problem graphically.
Math Aside
Existence of Unique Maximal Element: If a consumer faces a convex budget
set X over which he tries to maximize a CT, strictly monotone and strictly convex
preference relation then the maximal element is unique. To prove that, let us
suppose the maximal element is not unique. Then there exists x, y ∈ X such that
both x, y ∈ C(X), where C(X) denotes the set of maximal elements.
Since the budget set X is convex, any convex combination of x and y also belongs
to X. Consider a convex combination of x, y, namely, λx+(1−λ)y where λ ∈ (0, 1).
By strict convexity we get λx + (1 − λ)y x ∼ y and from convexity of budget set
we get λx + (1 − λ)y ∈ X. This violates the assumption that x, y ∈ C(X). Since we
have shown already that for a CT and continuous preference the maximal element
must exists, it has to be unique.
21
22 CHAPTER 3. DEMAND
Note that, when the agent is choosing optimally, i.e., at point x∗ the IC is
tangent to the budget constraint. If the IC is not tangent to the budget line and
intersects the budget line then we can find another IC in the NE direction which
gives higher utility but still remains feasible. Thus the condition for the optimal
solution is that the slope of the IC should be equal to the slope of the budget line,
i.e., the budget line and IC are tangent at the optimal bundle.
The optimal bundle chosen from a budget set thus depends on prices and income,
hence it is written as (x̂1 (p1 , p2 , m), x̂2 (p1 , p2 , m)), that is a function of price vector
and income. These functions are called demand function in economics since they
determine the quantity of a good demand for a set of given prices and income.
We are interested in analyzing the change in the optimal decision when income
changes. In the previous section we have already derived the optimal choice of the
agent, x̂1 (p1 , p2 , m) and x̂∗2 (p1 , p2 , m). For this entire section, we will consider the
impact of changing various parameters on the optimal choice of good 1 but the
analysis for good 2 would be identical. The question we are interested to answer
here is how does the optimal choice for good 1 changes when income m changes.
We assume that the prices of good 1 and 2 are fixed at p̄1 and p̄2 resp.
Graphically, we can see the effect of change in demand when income changes
with a shift of the budget line. So we find the optimal points, namely the point
of tangency between the IC and budget line as m changes. If we join the optimal
choices for each budget line, the curve we generate is known as income offer curve
(IOC) or income expansion path. This curve is drawn in our usual x1 − x2 plane
where the change of money income can be difficult to see.
Instead, if we are only interested in the relation between x̂1 and m, we can use
a diagram that plots x1 on x-axis and m on the y−axis. If we draw the relationship
x̂1 (p̄1 , p̄2 , m) for each m in this graph, the resulting curve is called the Engel curve.
The diagram below illustrates.
We would call a good normal if an increase in income increases the demand for
it. A good would be called an inferior good if an increase in income decreases the
demand for the good. Also, a normal is called a luxury if increase in income cause
more than proportionate increase in demand. This is just a convenient classification
scheme. Note that this classification is defined locally, for a given level of income
an increase in income can reduce demand but at a different level of income it can
increase demand as well. The following diagram shows the classification scheme
for good x1 . If the black line is the original budget line then the different regions
denote where the new optimal bundle will be after an increase in income.
The income effect for a luxury good and inferior good is illustrated below:
24 CHAPTER 3. DEMAND
optimal choice of good 1, x̂1 (p1 , p̄2 , m̄) on a price quantity (of good 1) diagram.
This would generate the demand curve for good 1. This is the usual demand curve
we are all familiar with. The diagram below illustrates.
Based on price effect we can divide goods into two categories, Giffen good and
ordinary good . If the demand for a good increase with an increase in price it is
called a Giffen good, otherwise an ordinary good. This is also a local classification
scheme.
the same. Since the price of good 2 does not change let m0 be the income that keeps
the old bundle just affordable. Then
The substitution effect is the change in the optimal choice of x1 as the price
ratio changes, keeping the real income same. This is capture by the movement of
the optimal choice from the original budget line to the hypothetical budget line.
The income effect on the other hand is the change in the optimal choice of x1 as the
real income changes. This would be captured by the movement of optimal choice
from the hypothetical to the new budget line. The diagram below illustrates,
In the diagram above the price p1 decreases from the red line to the black line.
The black dashed line represents the adjust budget line with money income m0 .
Let xold
1 and xnew
1 are the two optimum consumption levels of good 1 for the red
and black budget line resp. The movement from red line to the dashed black line
denotes the substitution effect and the movement from the black dashed line to
solid black line denotes income effect. Combining together, the movement from the
red line to the black line would give the total price effect.
Substitution effect always increases the demand when price decreases. It is easy
to prove it geometrically. Suppose not. Suppose after a drop in price of good 1
the DM chooses lower amount of x1 then as you can see in the diagram below, the
optimal choice will be on a lower IC. Since old choice is affordable at income m0
this cannot be optimal for the DM. In general the substitution effect ∆S has the
3.2. COMPARATIVE STATICS 27
opposite sign of that of the price change ∆p. In other words the substitution effect
∆S /∆p ≤ 0.
Even though the substitution effect is always negative the income effect can be
negative or positive. The earlier diagram denotes the case where the substitution
effect and income effect due to the decrease in the price of good 1 go in the same
direction but that may not always be the case. In the diagram below we show the
other case where the two effects go in the opposite direction. This happens when
good 1 is an inferior good.
This way the income effect component links the classification due to price change
28 CHAPTER 3. DEMAND
and due to income change. The sign of the income effect depends on whether a good
is normal or inferior. For normal goods, the income effect raises the optimal choice
of x1 as price decreases. But for inferior goods, as price decreases the real income
increases and the optimal choice for x1 decreases (as shown in 3.8).
The sign of the substitution effects leads to a well-known result in economics,
namely Fundamental theorem of consumption theory, or more commonly law of
demand. The theorem says, if the consumption of a good increase with the increase
in income then the consumption would decrease with the increase in own price. It
is apparent from the Slutsky decomposition.
The SE for cross price is always positive (how will you show that?) and for a normal
good the IE is always positive. If the SE > IE then the P E > 0, which means the
two goods are substitutes and if SE < IE, the P E < 0, the two goods are then
complements.
3.3 Examples
We will consider two extreme examples, perfect substitutes, perfect complements.
The diagrams below illustrate:
Perfect Substitutes
Slutsky decomposition can be different for different level of price change. The
diagram below illustrates. In the first case, since the change in price is sufficiently
the DM buys only x1 before and after the price change and IE = P E. In the second
case, however, at the old prices the DM was not buying any positive amount of x1
and a substantial change in prices imply that he is only buying x1 after the price
change, hence SE = P E.
3.3. EXAMPLES 29
Perfect Complements
For perfect complements the DM needs to buy both goods in the same ratio irre-
spective of prices. Thus there is no room for substitution which leads to IE = P E.
Utility
In the last chapter, we have defined preference relation over commodity bundles and
have used IC for a graphical representation. But, for most of the practical purposes,
we will need an algebraic representation of the preference, namely we want to write
a function involving the two goods that will tell us between two commodity bundles
which one is preferred. Such a function is called utility function.
31
32 CHAPTER 4. UTILITY
This is true because if x is a maximal element for % over X then x % y for all
y ∈ X, but the equivalence implies u(x) ≥ u(y) for all y ∈ X, thus x will also be
maximal according to the utility function. On the other hand if u(x) ≥ u(y) for all
y ∈ X, that is x is the maximal element according to the utility function u(.) then
x % y for all y ∈ X. Thus x is also a maximal element according to the preference
order %.
However, if the implication was one-sided then maximizing preference would not
be same as maximizing the utility function. Suppose,
x % y ⇒ u(x) ≥ u(y)
then consider the example where X = {x, y} with x % y and define u(x) = c
for some constant c for both x, y ∈ X. The optimal choice for % is x but since
u(x) = u(y) = c, maximizing utility would be lead to choice of x or y.
Math Aside
Existence of Utility Function
The necessary condition for the existence of utility representation for any preference
relationship % is completeness and transitivity. However, completeness and transi-
tivity is not sufficient for the existence of a utility representation. An example is a
lexicographic preference on R2+ does not have a utility representation even though
it satisfies completeness and transitivity.
Furthermore, if the utility function is continuous then the utility function is con-
tinuous as well. Continuity, as defined below guarantees the existence of the utility
function.
Continuity: A preference relation is continuous if the weakly preferred set of any
bundle is closed. In other words, if we consider a converging sequence of commodity
bundles namely, x1 , x2 · · · → x such that all of these bundles are preferred to y then
their limit is also preferred to y, i.e., x % y. This technical assumption ensures that
a maximal element exists for the choice set X ∈ R+ n.
4.2. MARGINAL UTILITY AND MRS 33
where ū denotes the amount of utility the consumer gets from the commodity
bundle. Then instead of writing in terms of commodity bundle, we can denote the
ICs in terms of ū they generate.
Now that we have written ICs in terms of the utility function, we can represent
the slope of the utility function algebraically. First consider how utility changes
when the amount of one good changes, this can found by looking at the partial
derivative of the utility function. The amount of change in utility due to change in
one good is called the marginal utility (MU) of that good,
∂u
M U1 = (MU)
∂x1
Now going back to the IC, if we take a total derivative of both sides of equation
IC, we get
∂u ∂u
dx1 + dx2 = 0
∂x1 ∂x2
by rearranging we get,
∂u
dx2 M U1
M RS = = − ∂x
∂u
1
=− . (4.1)
dx1 ∂x2
M U2
34 CHAPTER 4. UTILITY
subject to p1 x1 + p2 x2 = m
As in the previous chapter we will assume that the preference is strictly monotone
and convex. Strict monotonicity of % implies is u(.) represents % then for any two
vectors x̃, ỹ ∈ X ⊂ RN+ , if x̃ > ỹ then u(x̃) > u(ỹ). This implies for a standard
budget set the optimal choice will lie on the budget line, since otherwise a movement
in NE direction will increase utility.
The convexity of % implies that slope of IC decreases with an increase in x1 .
Optimal choice given an utility function will be obtained by equating the slope of
the IC and the budget line, as in the graphical method. This implies,
p1
M RS} = − (4.2)
| {z p
slope of IC |{z}2
slope of budget line
M U1 p1
= (4.3)
M U2 p2
The slope of the IC gives the ratio of the marginal benefit of good 1 against good 2
and the slope of the budget line gives the opportunity cost of good 1 against good 2.
Equation 4.3 says the marginal cost has to be equal to the marginal benefit which
is the fundamental condition for optimization.
However 4.3 is one equation in two variables x̂1 and x̂2 , the two optimal levels of
goods. To solve for then we need the budget lines that ensures two equations in two
unknowns. Solving them simultaneously gives us the solution of the constrained
optimization problem.The solution of the problem would be a function of the pa-
rameters, so we can write them as x̂1 (p1 , p2 , m) and x̂2 (p1 , p2 , m). The function
x̂1 (p1 , p2 , m) is called the Marshallian demand function for good 1 and similarly
x̂2 (p1 , p2 , m) is called the Marshallian demand function for good 2.
In the following two subsections we show two different methods of solving the
optimization problem in the two goods case that leads to the same solution.
Substitution Method
For this method we eliminate the constraint and substitute x2 by a function of x1 in
the objective function. Then we can solve the unconstrained optimization problem
for only x̂1 and recover x̂2 using the predefined substitution relationship. To do so
we rewrite the budget equation as follows:
m p1
x2 = − x1 . (4.4)
p2 p2
4.3. CHOICE PROBLEM: SOLVING ANALYTICALLY 35
The function U (p1 , p2 , m) is called the indirect utility function, since this is the
utility function in terms of money.
∂L (x1 , x2 , λ)
= u2 − λp2 = 0 (4.7)
∂x2
∂L (x1 , x2 , λ)
= m − p 1 x1 − p 2 x2 = 0 (4.8)
∂λ
We can rewrite equations 4.6 and 4.7 as
u1 = λp1 (4.9)
u2 = λp2 . (4.10)
Dividing equation 4.9 by 4.10 we get,
u1 p1
−M RS = =
u2 p2
Which is again same as before. If we combine this equation with 4.8, which is the
budget constraint, we can solve for the demand functions of x̂1 and x̂2 .
One of the main reason behind using the Lagrangian multiplier method is that
the Lagrangian multiplier has an economic interpretation. One can show that
∂L (x1 , x2 , λ) ∂U (p1 , p2 , m)
= = λ.
∂m ∂m
This implies λ is the amount of increase in the indirect utility if we increase
the income by the unit. In other words, this shows the opportunity cost of the
constrained and that is why it is called the shadow price of the constraint. If we
can relax the constraint by one unit we would get λ extra units of utility, thus λ
tells us how costly it is to have the constraint.
Finally combining all the methods together we can see that all methods lead
to the same conclusion namely, −M RS = price ratio. Each method has it’s own
advantage and disadvantage and depending on the context we may use one or the
other.
2. x∗ = 0 only if f 0 (0) ≤ 0
3. x∗ = X̄ only if f 0 (X̄) ≥ 0.
4.3. CHOICE PROBLEM: SOLVING ANALYTICALLY 37
Thus maximizing a concave either lead to a corner solution where x̂1 = 0 or x̂2 = 0
or the FOC is the sufficient condition for maximization.
The following two diagrams show the maximum element x̂ for a one dimensional
function f (x) defined over [0, B]. The diagram below shows the corner solutions,
The diagram below illustrates the interior solution,
One example of boundary solution in the standard two good case is that of
linear utility function u(x1 , x2 ) = ax1 + bx2 . The optimal solution for this linear
function given a standard budget line usually lies at the corner. The diagram below
illustrates.
In the above diagram the parallel blue lines represents the IC function and the
red line is the budget line. Since the budget line is flatter, i.e., the Marginal Cost
(MC) for consuming 1 additional unit of x1 is lower than the slope of IC, i.e.,
Marginal Benefit (MB) from consuming 1 additional unit of x1 , the consumer will
optimally choose to consume the corner, where x̂2 = 0.
Convex Preference and Concave Utility function For convex choice set X
and concave utility function the preference relationship is convex. To show that we
consider x, y ∈ X and λ ∈ [0, 1] such that
where u(.) represents the preference relationship % and x ∼ y. Since u(.) represents
%, this implies u(x) = u(y). Going back to the concavity condition of utility function
we get,
utility function the IC are bowed in, which gives us an unique maximal element in
the two-good case.
The problem with non-convex preference or non-concave utility is described in
the diagram below. In the diagram below at point, x̂1 the agent is on IC1 . Even
though the FOC is satisfied at x̂1 , as the MRS equals to the price ratio the SOC
fails here. If the agent moves to x̂2 , he is at a higher level of utility on IC2 and
at x̂2 also the budget constraint is satisfied. Thus x̂1 can not be the optimal or
utility-maximizing bundle.
good 1 is a luxury good then the income offer curve bends towards x-axis, i.e., good
1.
Note that, if we plug the optimal choice in the budget line we get,
Thus the weighted average of income elasticity adds up to 1, where the weights are
given by expenditure share, pm i x̂i
for i = 1, 2.
Next consider the impact of change is p1 on the optimal choice of x̂1 . Similar
to the analysis before this change is summarized by ∆x̂1 (p∆p1 ,p1 2 ,m) . The own price
elasticity, mostly written as price elasticity is defined in a similar manner as,
% change in demand p1 ∆x̂1 (p1 , p2 , m)
1 = price elasticity = = .
% change in own price x̂1 ∆p1
Continuing with the classification scheme if the price elasticity is positive, i.e,
as the price of good 1 decreases the agent would optimally choose to consume less
of good 1, we will define it as a Giffen good. Otherwise, we call it a ordinary good.
If we consider the effect of change in p2 on x1 we can capture it by ∆x)
∆p2
1
and the
corresponding cross price elasticity is given by,
% change in demand of good 1 p2 ∆x̂1 (p1 , p2 , m)
12 = cross price elasticity = = .
% change in price of good 2 x̂1 ∆p2
Similar to the graphical analysis, if the cross-price elasticity is positive, i.e., as the
price of good 2 increases demand for good 1 increases then the goods are called sub-
stitutes. If the cross-price elasticity is negative, i.e., consumption of good 1 increases
when the price of good 2 increases then the two goods are called complements. Note
that the earlier examples of perfect complements and perfect substitutes are special
cases of this complement and substitute goods.
We can also revisit the Slutsky Decomposition. As before, suppose price of good
1 reduces from p1 to q1 and p2 , m remains the same. Then the hypothetical income
that keeps the original bundle just affordable would be
Let x̃(q1 , p̄2 , m0 ) denote the chosen bundle with the new price ratio but same real
income. Then we can rewrite the impact of price change (Price Effect) as follows:
x̂1 (q1 , p̄2 , m)−x̂1 (p1 , p̄2 , m̄) = [x̂1 (q1 , p̄2 , m0 ) − x̂1 (p1 , p̄2 , m̄)] + [x̂1 (q1 , p̄2 , m) − x̂1 (q1 , p̄2 , m0 )]
| {z } | {z }
∆S ≡substitution effect ∆m ≡income effect
4.3.3 Examples:
Let us first revisit our examples from the earlier chapter an then using the algebraic
structure of the utility function, we will consider a few more interesting examples.
Math Aside
Homothetic Preference: A preference relation is such that when (x1 , x2 ) ∼
(y1 , y2 ) then for all t > 0 we get (tx1 , tx2 ) ∼ (ty1 , ty2 ), then we call the preference
relation is homothetic. The utility function also has the same property namely,
Since ICs shift parallely along a ray through the origin if there is a change in income
only the optimal choice will remain on the same ray, i.e., the income elasticity is
constant for homothetic preferences.
Next we will show the two methods of constrained optimization for a general two-
good Cobb-Douglas utility function.
rearranging,
m p1
αp2 − x1 = (1 − α)p1 x1
p2 p2
cross-multiplying we get,
αm = (αp1 + (1 − α)p1 ) x1
Lagrangian Multiplier method For the Cobb-Douglas utility function the La-
grangian function is given by,
L (x1 , x2 , λ) = α ln x1 + (1 − α) ln x2 + λ (m − p1 x1 − p2 x2 )
If we also plug the values of x̂1 and x̂2 in the utility function, we get the following
indirect utility function:
αm (1 − α)m
U (p1 , p2 , m) = α ln + (1 − α) ln (4.12)
p1 p2
∂U (p1 , p2 , m) α (1 − α) 1
= + = = λ∗ . (4.13)
∂m m m m
This shows that the Lagrangian multiplier is indeed the shadow price of relaxing
the constraint.
Note that, the Cobb-Douglas utility function is concave since,
N
X
u(x̃) = αn ln xn
n=1
is a sum of N concave functions. Thus we do not need to check for the sufficiency
condition.
Given the demand function for good 1 for the Cobb-Douglas utility function,
namely,
m
x̂1 = α
p1
m ∂x1
η1 = = 1.
x1 ∂m
p1 ∂x1
1 = = −1
x1 ∂p1
p2 ∂x1
12 = = 0.
x1 ∂p2
The same result holds true for demand for good 2 as well. This generates an
interesting feature in the Slutsky decomposition diagram. If the price of good 1
changes the new bundle would be on the same horizontal line as the old bundle and
if the price of good 2 changes the new bundle would be on the same vertical line as
the old bundle. So for the cross price the SE and the IE cancels out.
4.3. CHOICE PROBLEM: SOLVING ANALYTICALLY 45
Note that all the elasticities are constant for Cobb-Douglas utility function.
Cobb-Douglas utility function belongs to a class of utility function known as Con-
stant Elasticity of Substitution (CES) utility function. The name is derived from
the fact the price elasticities are constant for these type of utility function. We can
write the general two-good CES utility function as
Quasi-linear preference:
u (x1 , x2 ) = v (x1 ) + x2
where v (x) can be any concave function of x1 . Note that the function is linear in
x2 , the composite good but not in good 1. That is why this is called a quasi-linear
preference. An interesting feature about the IC for the quasi-linear good is that
they are vertical shifts on each other.
46 CHAPTER 4. UTILITY
Since v(x1 ) is concave the quasi-linear utility function is also concave. This
implies for the quasi-linear utility the sufficiency condition is also satisfied.
Perfect Substitutes:
For the perfect substitutes case we showed that the IC is a downward sloping line,
thus the utility function can be written as
u (x1 , x2 ) = ax1 + bx2 ,
where − ab denotes the slope of the IC which is also the MRS between the two
goods. Let us consider an example where ū = 5 and a = 1, b = 5. Then the agent
would be indifferent between 5 units of good 1 and 1 unit of good 2. Thus the
MRS denotes the rate at which the agent is willing to trade good 1 for good 2. For
perfect substitutes, the MRS remains constant. Given the MRS − ab if the price
ratio is such that pp12 < ab then the consumer will only consume x1 and no x2 and if
p1
p2
> ab then he will consume only x2 , thus the demand function for good 1 is given
by,
m p1 a
p1 if p2 < b
x̂1 (p1 , p2 , m) = [0, pm1 ] if pp21 = ab
0 if pp21 > ab
The price elasticity depends on the level of change in price of good 1. We will only
consider one case, pp21 < ab where the agent is consuming only good 1. If the price
p01 a
change for good 1 is such that p2
< b
then
p1 ∂x1
price elasticity = = −1
x1 ∂p1
4.3. CHOICE PROBLEM: SOLVING ANALYTICALLY 47
p0
but if the change in price of good 1 is such that p12 > ab , then the consumption of
good 1 goes to zero as the consumer only starts to consume good 2. We can do
similar exercise for price of good 2.
Perfect Complements:
The perfect complements case is the polar opposite of perfect substitutes. In case of
perfect complements the agent would consume the two goods together. For example
suppose our consumer only takes one cup of coffee with two spoons of sugar. That
means he only consumes (1C, 2S) at any time. More sugar without coffee is not
valuable for him. More coffee without sugar is also not valuable to him. Thus we
can write the utility function as
The IC would have a kink at a point where ax1 = bx2 . (Think about the slope of
the line that joins the kinks for different IC and the origin!) The reason we get the
L-shaped IC is the following; if ax1 > bx2 , then u(x1 , x2 ) = bx2 which implies in
the region where ax1 > bx2 the IC is given by bx2 = ū, which is a horizontal line
at x2 = ū/b. For the region where ax1 < bx2 we have u(x1 , x2 ) = ax1 and the IC
would be given ax1 = ū, which is a vertical lien at x1 = ū/a. The vertical and the
horizontal line forms the two hands on the L-shaped IC. Note that, the optimal
bundle will always be on the kink (think about what happens if that is not the
case), we can get the consumption bundle at the intersection between the budget
line and kink. This implies the demand function for good 1 is given by,
m
x(p1 , p2 , m) =
p1 + p2
Then the income elasticity is given by,
m ∂x1
income elasticity = =1
x1 ∂m
and the price elasticity is given by,
p1 ∂x1 p1
price elasticity = =−
x1 ∂p1 p1 + p2
and the cross price elasticity is given by,
p2 ∂x1 p2
cross price elasticity = =− .
x1 ∂p2 p1 + p2
48 CHAPTER 4. UTILITY
Chapter 5
Revealed Preference
In the previous chapters, we have followed two approaches to solve for the optimal
choice problem of the decision maker. In the first one, we have taken the preference
as the primitive and then derived the choice of the agent by combining the prefer-
ence with the budget constraint. In the second approach, we have defined utility
functions that represents a preference and found the optimal choice using standard
calculus technique.
In this entire analysis, our main assumption was agents are rational, they are
maximizing their gains subject to the constraint they face. But how can we be sure
whether that is true? If we could somehow know the preference of an agent then
we could test whether they are maximizing or not. But the concept of preference
is in the mind of the consumer and it is impossible to know for an economist. We
can try to test different assumptions we have made on this preference relation, but
that can also be computationally difficult. For example, suppose you have a choice
set with 50 elements, in order to check for completeness, we need to test all 50 2
pairs to test that. For a standard budget set this implies checking infinite pairs of
elements which is impossible. Thus it is impossible to operationalize the theory of
choice using preference relationship.
Note that, the only thing we observe about the consumers is their final choice,
i.e., how much of each good the agents are choosing given the prices. Thus any
theory of testable implication will only rely on the observable choice data. In this
chapter, we will build a framework that just takes the choice data as given and
generates a theory of optimal choice based on some assumptions on the choice data.
A choice data refers to the data set consisting of the tuple of price, income and
chosen bundles. For example, suppose X = RiN and there are J price, income pairs
are available then the choice data is given
where each x̄j and p̄j are N −dimensional vectors and all prices and income are
strictly positive.
Our goal is to find conditions under which the choice data represents a maximiza-
tion over a standard monotonic utility function. Before we impose any condition
on the choice data, to ensure monotonicity it must be the case that at the chosen
bundle x̃j at price p̃j we get
p̃j .x̃j = mj .
49
50 CHAPTER 5. REVEALED PREFERENCE
To form a theory of choice we need two conditions on the data set. First, what
conditions will be implied by the standard utility representation and second what
conditions will violate the standard utility representation. In the rest of the chapter
we will impose conditions to build the ”if and only if” relationship between choice
data and the standard utility representation. This means if x̃j is chosen at price,
income pair (p̃j , mj ) then we want to find conditions under which x̃j is the optimal
choice from the following budget set
B j = z̃ ∈ RN j j
+ |p̃ .z̃ ≤ m .
All the elements z̃ is the budget set B j are affordable when x̃j is chosen. Using
this notion of affordability let us define a new binary relationship namely, revealed
preferred to on our choice data. For a given price p̃j if bundle x̃j when income level
is mj then we say x̃j isrevealed preferred to bundle z̃ if z̃ is strictly affordable at p̃j
and mj ,
For a standard utility representation we need if x̃RP z̃ then u(x̃) > u(z̃).
Consider the choice data set with J = 2, p̃1 = (1, 2), x̃1 = (3, 1), m1 = 5 and
p̃2 = (2, 1), x̃2 = (1, 3), m2 = 5. At p̃1 , x̃2 is not affordable since
and similarly x̃1 is not affordable at p̃2 so none of them are revealed preferred to
the other.
Contrast this with the example of a choice set where p̃1 = (1, 2), x̃1 = (1, 3), m1 =
7 and p̃2 = (2, 1), x̃2 = (3, 1), m2 = 7. Here, at price p̃1 the bundle x̃2 is affordable,
and vice versa. So x̃1 is revealed preferred to x̃2 and vice versa.
In the second example, it is clear that a consumer with a monotone utility
function is not choosing optimally. Inspired by this example we introduce the first
axioms of this chapter.
The left panel denotes a case where W ARP is not violated as in the case with
the first example and the second diagram shows the case where W ARP is violated
similar to example 2.
Now consider the third example :
In this example x̃1 RP x̃2 x̃2 RP x̃3 and x̃3 RP x̃1 ,i.e., it generates a cycle of re-
vealed preferred relationship even though WARP is satisfied. Hence, we need to
strengthen our assumption to rule out circularity of choices. Otherwise, according
to a standard utility function we will get
x1 x2 x3
p1 5∗ → 4↓ 6
p2 ↑6 5∗ → 4↓
p3 ↑4 ←6 ← 5∗
Start with a bundle on the diagonal. If there exists another bundle with a lower
expenditure in the same row then that bundle is affordable at the row price. For
52 CHAPTER 5. REVEALED PREFERENCE
this bundle go vertically to the diagonal element and repeat. If a cycle exists then
SARP is violated. No cycle would imply SARP is satisfied.
WARP is a necessary condition for the existence of a continuous and monotone
utility function. SARP is stronger since it implies the sufficiency. If SARP is
satisfied for a finite number of commodity bundles then a continuous, monotone and
concave(convex IC) utility function exists. Thus the existence of utility function is
equivalent to checking SARP on finitely many commodity bundles which is much
easier than recovering the preferences from the consumer’s mind. This theory thus
has a strong influence on empirical approach that tries to recover the consumer
demand function based on choice data.
Note, this again leads to the famous Fundamental Theorem of Consumption
Theory more commonly known as Law of Demand in economics. To recall, the
theorem says that if the demand for a good increases when income increases then
demand decreases for an increase in price. We will explore this further in the
problem set.