Preference and Utility I: Rational choice

Lam, Wing Tung

ECON2210 Fall 2024

ˆ A central question in Economics is how an individual makes a decision. We model

the situation as the individual chooses one out of multiple options. We begin with
the case that she is directly concerned about the option chosen. We abstract away
interactions with other entities and future decisions.

ˆ A fundamental premise of Economics is that a “rational” individual always makes

an “optimal” decision. The first part of the notes formalizes the above notion. We
introduce the concept of rational preference.1 It allows us to define the individual’s
optimal option as her most preferred option in the feasible set, the set of available
options.2

ˆ The second part introduces the concept of utility and its connection to rational
preference. We explain why an individual makes her choice as if she maximizes
her utility.

ˆ The utility maximization framework serves as the foundation for what we can learn
from observing past choices made by the individual and what we can predict about
individual’s choices in various circumstances. It is formally known as the theory of
rational choice. We then discuss the properties of optimal choices, which form a set
of behavioural restrictions.

ˆ Lastly, we connect the concept of utility to welfare and opportunity cost.

ˆ From a methodological standpoint, this notes provides an introduction to economic

modeling.
1
In all notes, quotation marks indicates that the word is used in a layman and informal manner, while
italic font indicates that the wording is a formal terminology.
2
Some textbooks call feasible set as choice set.

1
 – The abstraction highlights the key elements and the generality of the analysis.
As the general setting applies to any institution and many applications, so do
our conclusions.
– The formalism spells out all underlying assumptions and leaves no room for
ambiguity. This distinguishes the formal terminologies from the layman language.

1 Preference
ˆ We first build a mathematical model for decision-making.

ˆ An individual must choose one option. There are multiple options which an individual
may potentially choose from. Depending on the application, an option can be
anything that directly concerns the individual. Examples are

– “pasta” or “burger” when deciding supper

– “Maldives” or “Madrid” when deciding vacation destination
– Consumption bundles consists of multiple goods when making consumption
decision

Example. Tom is treated to a double scoop at a gelato booth. Tom picks the
flavors of the two scoops from coffee (C), sesame (S) and vanilla (V).

Here an option specifies the flavors of the two scoops e.g., (C, S) denotes a coffee
scoop and a sesame scoop. Even though Tom chooses up to two flavors at once,
he is indeed choosing ONE option from the following:

{(C, C), (C, S), (C, V ), (S, S), (S, V ), (V, V )}

ˆ We denote a generic option using an upper case letter, e.g. X or Y

ˆ We model a household’s preference using the preference relation ⪰. ⪰ describes the

weak preference between any two potentially feasible options.

ˆ X ⪰ Y is read as “household weakly prefers X to Y ” or “X is at least as good as

Y ”. It is equivalent to write Y ⪯ X.

2
 ˆ We can further divide weak preference relation into strict preference and indifference
relations

ˆ X ≻ Y is read as “household strictly prefers X to Y ” or “X is strictly better than

Y ”. We define the strict preference relation ≻ using ⪰

X ≻ Y if X ⪰ Y and NOT Y ⪰ X.

– By definition, X ≻ Y implies X ⪰ Y .3
– Suppose only X and Y are feasible, X is always chosen if X ≻ Y .

ˆ X ∼ Y is read as “household is indifferent between X and Y ” or “X and Y are

equally preferred.” The indifference relation ∼ is defined using ⪰

X ∼ Y if X ⪰ Y and Y ⪰ X.

– X ∼ Y implies X ⪰ Y .
– Suppose only X and Y are feasible, either options may be chosen if X ∼ Y .

ˆ The above definitions imply that if X ⪰ Y , then either X ≻ Y or X ∼ Y .

ˆ Do not confuse symbols ⪰ and ≻ with inequalities ≥ and >. ⪰ and ≻ apply to
options. ≥ and > only apply to numbers.

ˆ Traditionally, economists simply take preference as a primitive object in the model.

– Economists are agnostic about how a household’s preference is formed and do

not study the psychological and biological determinants of preference.

ˆ We now introduce two fundamental assumptions on the preference relation

Definition (Completeness). A preference is complete if for any two options

X and Y , we have either X ⪰ Y , Y ⪰ X or both.

3
If you find the preceding statement counter-intuitive, you may consider the contrapositive of the
statement: if X ⪰ Y does not hold, then X ≻ Y will not hold. This is because X ≻ Y requires both
X ⪰ Y and NOT Y ⪰ X.

3
 ˆ The definition of completeness is equivalent to the following: For any two options X
and Y , exactly one of the three cases {X ≻ Y , X ∼ Y , Y ≻ X} must be true.4

ˆ Preference is an INVARIANT object in the model framework. The notion of completeness

not only requires household to be able to compare any two options, but have the same
ranking irrespective of what other options are feasible.

– In words, the household ranks any two options in the same way over time and
in all circumstances!

Example (The Fox and the Grapes). A fox describes its preference as follows:

– If he can jump highly enough to grab the grapes from the trees, he strictly
prefers grapes to blueberries.
– If he cannot reach the grapes, he goes away sour and strictly prefers
blueberries to grapes.

The fox’s preference is incomplete. Its preference over blueberries and grapes
depends on whether the grapes is in its feasible set!

ˆ To have a complete preference, economists incorporate relevant attributes of the

circumstance, or so-called state, in the definition of an option.

– For example, “having hotpot with friends in a cold environment” and “having
hotpot alone in a warm environment” are different in a household’s perspective.
We should then define them as two separate options. This ensures the same
ranking of options amid changes in circumstances e.g., weather and occasion.

Definition (Transitivity). A preference is transitive if for any three options

X, Y and Z,

X ⪰ Y and Y ⪰ Z imply X ⪰ Z.

4
Some textbooks split completeness into two properties. They define completeness over two
DIFFERENT options, and introduce the notion of reflexivity X ⪰ X for the case X = Y .

4
 – Transitivity rules out “cycles” in the preference relation e.g., X ⪰ Y and Y ⪰ Z
but Z ≻ X.

Assumption (Rational preference). A household’s preference is complete

and transitive. We say such a preference is rational.

– From now on, we consider only rational preferences, unless specified otherwise.

ˆ We imposed transitivity on the weak preference relation. The strict preference and
indifference relations inherit transitivity by definitions. The remark below provides
some examples

Remark. Suppose the preference is complete and transitive. The following

statements hold:

– If X ∼ Y and Y ∼ Z, then X ∼ Z.
– If X ⪰ Y and Y ∼ Z, then X ⪰ Z.
– If X ≻ Y and Y ∼ Z, then X ≻ Z.
– If X ≻ Y and Y ⪰ Z, then X ≻ Z.

– In appendix, we prove the remark using the formal definitions and give you a
glimpse of the work done behind the scenes.5
– Bear in mind that the formal definitions of preference relations are not the same
as “preference” in our daily usage. The latter is the interpretation of the model.
Do not confuse interpretations with definitions.
– Take a step back. We introduced a mathematical symbol ⪰ with two properties:
completeness and transitivity. We then defined two more symbols ≻ and ∼. We
claimed that these symbols form a model capturing a household’s “preference”
in our daily usage. Does such an interpretation make sense? The remark
indicates that the model implications are in line with the intuitive properties of
“preference” at a first pass.
5
Materials in appendix will not be assessed.

5
Example (Condorcet voting paradox). Consider a committee consists of three
members {Andy, Bonnie, Cindy}, which purports to decide between TWO
options each time. The committee decision is determined by plurality voting,
the option with most votes gets selected. There are three options A, B, and C in
total.

We define the committee’s preference ⪰Com based on the voting outcome:

Suppose two options X and Y are put to vote. If committee always selects X,
then X ≻Com Y. If committee sometimes chooses X and sometimes chooses Y ,
then X ∼Com Y .

Every committee member’s preference is RATIONAL. They sincerely vote for

their own preferred option.

– Andy’s preference is A ≻ B ≻ C. It means A ≻ B, B ≻ C, and A ≻ C for

rational preference.
– Bonnie’s preference is B ≻ C ≻ A.
– Cindy’s preference is C ≻ A ≻ B.

We then derive committee’s preference from its members’ preference. When

options A and B are put to vote, Andy and Cindy vote for A while Bonnie votes
for B. Then the committee selects A, so we obtain A ≻Com B. When B and C
are put to vote, Andy and Bonnie vote for B, so B ≻Com C. When A and C are
put to vote, Bonnie and Cindy vote for C, so C ≻Com A.
The committee’s preference is given by

A ≻Com B; B ≻Com C; C ≻Com A

⪰Com is complete as committee has strict preference over any pair of options.

However, ⪰Com is intransitive! Since A ≻Com B and B ≻Com C, transitivity

requires A ≻Com C. This contradicts C ≻Com A.

Even though all individual members have rational preferences, the collective
choice of a democratic organization needs not be consistent with a rational
preference!

6
2 Utility
ˆ Utility is an alternative way to represent a household’s preference

ˆ A utility function U (.) assigns each potentially feasible option a number. The number
U (X) assigned to option X is called utility level of X

Definition. A utility function U represents a preference relation ⪰ if for any

two options X and Y ,

U (X) ≥ U (Y ) if and only if X ⪰ Y

ˆ In words, the more preferred options is always assigned a higher utility level, and
VICE VERSA.6

ˆ The following are also equivalent definitions of utility function

– U (X) > U (Y ) if X ≻ Y and U (X) = U (Y ) if X ∼ Y ;

– X ≻ Y if U (X) > U (Y ) and X ∼ Y if U (X) = U (Y ).

Exercise. Suppose there are three options A, B and C. Andy’s preference is

A ≻ B; B ≻ C ; A ≻ C

Or simply, A ≻ B ≻ C. The following table are the utility levels assigned by

four different utility functions. Which of the utility function(s) represent Andy’s
preference?

A B C
U1 6 2 1
U2 2 6 7
U3 4 2 0
U4 -1 -2 -3

6
To avoid confusion, you should see the definition as two separate statements, one with if and one with
only if. The if part says a more preferred option is assigned a higher utility level while the only if part says
an option with a higher utility level is more preferred.

7
ˆ Notice that a household’s preference can be represented by infinitely many different
utility functions, whereas her preference relation ⪰ must be unique.

ˆ Utility is an ordinal object. The utility level orders the options from the most
preferred to the least preferred. This order is the only useful information a utility
function contains

– Absolute levels of utility assigned to options have no economic meaning. Consequently,

the difference in utility levels across options has no economic meaning.

ˆ A key implication is that comparison of ordinal utility levels across individuals is

meaningless, so is summation of utility levels across individuals

Exercise (continued). Suppose there are three options A, B and C. Betty’s

preference can be represented by

A B C
UB 1 0 1

Andy’s preference is given by

A ≻ B; B ≻ C ; A ≻ C

The following statements compare Andy’s and Betty’s utility levels. For each of
the following statements, write down two utility functions representing Andy’s
preference, one utility function satisfies the statement and one utility function
violates it.

i “Andy assigns a higher utility level to A than Betty”

ii “Andy has lower utility levels on all options than Betty”
iii “Andy’s utility level from C is only half of Betty’s”
iv “When switching from B to A, the increase in Andy’s utility level is more
than twice the increase in Betty’s utility level.”

The above examples illustrate that any statement comparing different individuals’
utility levels is meaningless.

8
ˆ With a large number of options, it is easier to work with a utility function than the
preference relation ⪰

– Suppose there are M ≥ 2 options in total

– A utility function is comprised of M numbers, utility levels of the options.
M (M −1)
– A complete preference relation ⪰ is defined over C2M = 2 pairs of options
– C2M increases with M at a much faster rate! Therefore, utility function is a
more parsimonious object

ˆ From now on, we mostly work with utility functions. You may not find it unintuitive
that a household, including yourself, assigns a utility level to every option. It turns
out that this assumption is a consequence of rational preference.

Theorem (Utility representation). A preference defined over a finite number

of options can be represented by a utility function if and only if it is rational.

ˆ The only if part states that if we start with a utility function and write out the
preference it represents, that preference must be rational.

– The underlying preference must be complete because utility levels are a set of
invariant numbers which you can compare.
* For any two options X and Y , the utility levels U (X) and U (Y ) are the same
for any feasible set. They must satisfy either U (X) > U (Y ), U (X) = U (Y )
or U (X) < U (Y ).
– For transitivity,

X ⪰ Y and Y ⪰ Z
⇒ U (X) ≥ U (Y ) and U (Y ) ≥ U (Z)
⇒ U (X) ≥ U (Z) ⇒ X ⪰ Z

ˆ The only if part reminds us that when working with a utility function, we always
unconsciously use the completeness and transitivity properties embedded in the
utility function

9
 ˆ The if part states that transitivity and completeness together are the sufficient
conditions for a preference to be represented by a utility function.

ˆ The gist of the if part is that given a rational preference, we can always rank the
options from the most preferred to the least preferred and assign them utility levels
in descending order.

– In appendix, we describe a procedure on how to rank the options given a rational

preference.7

Example (Condorcet voting paradox continued). We cannot represent the

following committee’s preference using a utility function because it is not
transitive.

A ≻Com B; B ≻Com C; C ≻Com A

In other words, we cannot define committee’s “most preferred option(s)” and

“least preferred option(s)” in any meaningful way! A democratic society can be
seen as a large committee, and the notion “public interest” needs not be well-
defined.

3 Theory of rational choice

ˆ Now suppose only a subset of the options is feasible to the household. We call it the
feasible set.

ˆ The feasible set depends on the external environment. It changes with the circumstances.
Examples are

– List of candidates on a ballot depends on entry and drop-out

– Set of consumption bundles the household can afford in the market changes with
prices and budget.
7
Materials in appendix will not be assessed. The procedure can be generalized to construct utility
functions over infinitely many options. It is beyond the scope of this module.

10
 Example (continued). Tom is treated to a double scoop . Tom picks the flavors
of the two scoops at a gelato booth. Suppose the vanilla flavor (V) is sold out,
only coffee (C) and sesame (S) are available. The feasible set then becomes

{(C, C), (C, S), (S, S)}

ˆ For a given feasible set, we say a feasible option X e ⪰ Y for any other
e is optimal if X
feasible options Y .

– In words, X
e is a most preferred option in the feasible set. You cannot find a
strictly better option in the feasible set.8
– There can be more than one optimal options. Any one of the optimal options
may be chosen.

ˆ By the definition of utility function, Xe is an optimal option in a feasible set if and

only if X
e delivers the highest utility level among all feasible options

– It is much more convenient to maximize a numerical function than comparing

options pair by pair.

ˆ Economists assert that given any feasible set, the option chosen is optimal. The
notion of utility maximization is really an arcane way to say that the household
always picks her most preferred option in the feasible set

– We interpret that the household attains the utility level of her chosen option.
Then the household’s choice maximizes her utility level.

ˆ The if part of the utility representation theorem implies the following: Suppose a
household’s preference is rational. Her choices can always be analyzed in the way as
if she is maximizing some utility function. Empirically, we can always find a utility
function of which the maximization is compatible with her choices we observed. We
say such a utility function rationalizes her choices
8
Throughout the notes, ye denotes optimal value of variable y

11
 – Put it differently, we cannot model a household’s choices as maximizing a utility
function only if her preference is not rational

ˆ The usage of an economic model is threefold.

1. It predicts an outcome given a particular model specification. E.g., suppose

we know the exact utility function, we can find the individual’s choice in any
feasible set.
2. It translates the assumptions or properties of the model specification into restrictions
on the set of possible outcomes. E.g., we do not know the preference but assume
it is rational, we look for the restrictions on the individual’s choice.
3. It translates the past observations into information on the model specification,
and consequently further restrictions on the future outcomes. E.g., we do not
know the preference. The choices made by the individual informs us about how
she ranks the options.

ˆ Utility maximization imposes restrictions on how the set of optimal options CHANGES
with the feasible set.9 Let X
e be an optimal option of a feasible set.

– When other options are removed from feasible set, X e remains an optimal option
of the reduced feasible set. It is because X
e is still preferred to the remaining
options.
– When a new option Z is included into feasible set, the previously unchosen
options must remain sub-optimal. X
e is no longer optimal if and only if Z
is strictly better than X.
e In this case, Z becomes the optimal option of the
expanded feasible set.

Exercise. There are four options A, B, C, D.

1. Suppose that Susan’s preference is represented by

A B C D
US 4 2 0 5
Find her optimal option(s) in the following feasible sets.

9
The full set of restrictions is formally known as weak axiom of revealed preference or choice coherence.

12
 (a) {A, B, C, D}
(b) {A, B, C}
(c) {A, B}
2. Suppose that Walter’s preference is represented by
A B C D
UW 0 -1 2 2
Find his optimal option(s) in the above feasible sets.

ˆ Economists are interested in the restrictions on how the optimal option(s) varies
with the feasible sets. We call theses restrictions positive implications, or testable
implications, of the theory because they can be put to empirical tests.

ˆ The positive implications also allow extrapolation to choices we have never observed.

Example (continued). Suppose we do not know Patrick’s preference over the four
options A, B, C, D. But we observe that Patrick always chooses A from the set
{A, B, C}. From Patrick’s choice, we deduce that U (A) > U (B) and U (A) >
U (C).10

Suppose only A and B are feasible. We already know that U (A) > U (B). So
Patrick’s choice must be A.

Suppose all options are feasible. We do not know whether U (A) or U (D) is
higher. But we know for sure that B and C will not be chosen. We predict that
Patrick’s choice is either A or D.

Now suppose we observe that Patrick chooses C from the set {C, D} once. We
deduce that U (D) ≤ U (C) < U (A). We then modify our prediction. Patrick will
choose A when all options are feasible.

10
One may wonder why we take U (A) > U (B) and U (A) > U (C) instead of U (A) ≥ U (B) and U (A) ≥
U (C). This is to rule out the trivial preference that Patrick are indifferent between all options. In this case,
all feasible options are optimal. Such a preference can explain whatever choice Patrick makes. It also yields
no predictive power because we cannot rule out any option in a feasible set. Therefore, the convention is
to assume that unchosen options are strictly worse than the observed choice.

13
 – Stronger restrictions on the individual’ behaviour if we observe more of her
choices.

Exercise. “Utility is a metaphysical concept of impregnable circularity; utility is

the quality in commodities that makes individuals want to buy them, and the
fact that individuals want to buy commodities shows that they have utility.”
–Joan Robinson(1964)
Do you agree with the above statement? Is it true that the notion of utility
maximization is circular logic and not falsifiable ?

ˆ Suppose two situations present the same feasible set to a household. An even more
fundamental restriction of the utility maximization framework is that the household
will choose the same optimal option(s) in both situations. The appendix discusses
empirical evidence against such predictions.

Example (Condorcet voting paradox continued ). Recall that we derived

committee’s preference from its members’ preference

A ≻Com B; B ≻Com C; C ≻Com A

Here are other undesirable implications of the intransitivity:

– Suppose we do not know exactly the individual members’ preferences, but

we know they are rational and not indifferent between any two options.

If we observe a member’s votes in A vs. B and B vs. C, we may be able to

predict his vote in A vs. C. However, if the committee members vote using
secret ballot and we only observe the voting outcomes for A vs. B and B
vs. C, then no predictions can be made about the voting outcome for A vs.
C.
– The sequence of the pair of options presented will affect the final choice.
Say option A is status quo. Consider two different procedures.

14
 1. Other options B and C are first put to vote. Then the committee
chooses between the winning option and status quo A. In this procedure,
option A will be eventually chosen.
2. Committee votes between status quo A and other options one at a time.
If A is defeated in one of the votes, then the committee will choose the
winning option. If B and C both defeat A in the respective votes, B
and C will then be put to vote. In this procedure, the committee will
choose C at the end.

ˆ The theory of rational choice also has a normative appeal, prescribing how people
“ought to” make decisions. We probably will not upset a person if we comment her
preference as rational.

ˆ The primary interest in economics is to make predictions on what decisions people

actually make, not what decisions they ought to make. We will not focus on the
normative aspect.

Utility and welfare

ˆ Given a feasible set of options, we can find out the maximal utility level a household
may attain. It is exactly the utility level assigned to an optimal option.

ˆ We measure a household’s welfare in different circumstances using her maximal utility

level.

ˆ We say a household is better off with feasible set I than with feasible set II if her
maximal utility level is higher with feasible set I. In this case, the household is also
said to be worse off with feasible set II.

ˆ The household is better off under feasible set I in the following sense: if she is given
a choice between the two feasible sets, she always chooses feasible set I. This notion
of “better off” is testable!

ˆ Do not confuse the concept utility level with subjective feelings like “satisfaction”,
“happiness” or “well-being” in layman usage.

15
 – Change in subjective feelings is not measurable. Therefore, economists evaluate
a household’s welfare based on her choices, not subjective feelings.

Example. There are four options A, B, C, D. Suppose that Susan’s preference

can be represented by

A B C D
U 4 2 0 5

Initially, she can choose from options {A, B, C}, Susan will attain a utility level
of U (A) = 4

Now suppose Susan may replace option C with option D. The maximal utility
level under the feasible set {A, B, D} is U (D) = 5. Susan is better off with the
new feasible set.

Exercise. Can a household be worse off if she can choose from more options?

ˆ Utility maximization is often confused with self-interest. When we say an individual

maximizes her utility, it may give the false impression that she only cares about herself
but not others at all. The confusion stems from the misconception that utility level
is a measure of “well-being” or “happiness.”

– Though self-interest is a very common assumption in economic models, it is not

the same as utility maximization

ˆ Altruistic behavior, and other pro-social behavior, can be modeled using the utility
maximization framework!

Example. Suppose you have two pieces of custard buns with you and pass by a
homeless man in your neighbourhood. We define an option as (co , ch ) where co is
the quantity of buns you consume on you own and ch is the quantity given to the
homeless. For simplicity, let’s say the custard buns are indivisible, such that you
cannot split them into smaller portions. There are three feasible options: (2, 0),

16
 (1, 1) and (0, 2).
An example of utility function which captures altruistic behavior is the following:

U (1, 1) = 2; U (2, 0) = 1; U (0, 2) = 0;

In particular, sharing the custard buns equally with the homeless yields the
highest utility among all feasible options.

Exercise (continued). Propose a utility function U (co , ch ) which captures self-

interest.

Opportunity cost

ˆ The opportunity cost of choosing an option is the (utility level of the) most preferred
alternative option forgone.

ˆ An option is optimal if its utility level, or benefit, is weakly greater than its opportunity
cost.

– Utility maximization leads to minimizing the opportunity cost.

Example (continued). There are four options A, B, C, D. Suppose that Susan’s

preference can be represented by

A B C D
U 4 2 0 5

1. Consider the feasible set is {A, B, C}. A is the optimal option with U (A) =
4. The opportunity cost of choosing A is U (B) = 2, the next best option
in the feasible set. The opportunity cost of choosing B or C is U (A) = 4,
which is higher than their utility levels.

17
 2. Consider the feasible set is {A, B, D}. The optimal option becomes D. The
opportunity cost of choosing D is U (A) = 4 < U (D). The opportunity cost
of choosing A or B is U (D) = 5 which is above their utility levels.

ˆ The opportunity cost of an option depends on the feasible set of options.

ˆ When the feasible set changes, an increase in opportunity cost does not necessarily
make the household worse off.

– In particular, suppose new options are made feasible. The household may switch
to a new option to attain a higher utility level. At the same time, the opportunity
cost of an existing option can only be higher.

Example (continued). Initially, only A and C are feasible. Susan selects option
A, which opportunity cost is U (C) = 0

Suppose we expand the feasible set to {A, B, C}. Susan’s choice and utility level
are the same as before, while the opportunity cost of choosing A increases to
U (B) = 2.

Now option D is also made feasible. The opportunity cost of choosing A further
increases to U (D) = 5. Susan switches her choice from A to D, attaining a higher
utility level.

4 Key learning outcomes

ˆ Concept: Preference and utility, and their relation

ˆ Theory: Foundation and implications of utility maximization

ˆ Methodology: Usage of economic models

ˆ Skill: Comprehending formal statements

ˆ Skill: Distinction between formal terminologies and layman usage

18
A Appendix
A.1 Proof for properties of rational preference
ˆ The following examples show you the work economists did behind the scenes

Remark. Suppose the preference is complete and transitive. The following

statements hold:

– If X ∼ Y and Y ∼ Z, then X ∼ Z.
– If X ⪰ Y and Y ∼ Z, then X ⪰ Z.
– If X ≻ Y and Y ∼ Z, then X ≻ Z.
– If X ≻ Y and Y ⪰ Z, then X ≻ Z.

Example. Show that for any complete and transitive preference,

If X ∼ Y and Y ∼ Z, then X ∼ Z

To show X ∼ Z, we need both X ⪰ Z and Z ⪰ X.

X ∼ Y ⇒ X ⪰ Y and Y ∼ Z ⇒ Y ⪰ Z. From the definition of transitivity, we
obtain X ⪰ Z.
X ∼ Y ⇒ Y ⪰ X and Y ∼ Z ⇒ Z ⪰ Y . By transitivity, we obtain Z ⪰ X.

Exercise. Show that for any complete and transitive preference,

If X ⪰ Y and Y ∼ Z, then X ⪰ Z.

Example. Show that for any complete and transitive preference,

if X ≻ Y and Y ∼ Z, then X ≻ Z

To show X ≻ Z, we need X ⪰ Z and not X ⪯ Z.

19
 X ≻ Y ⇒ X ⪰ Y and Y ∼ Z ⇒ Y ⪰ Z. By transitivity, we obtain X ⪰ Z

It remains to rule out the possibility X ⪯ Z. Our approach is called proof by

contradiction, or reductio ad absurdum in logic. Suppose, to the contrary, that
X ⪯ Z. Together with Y ⪰ Z, we obtain Y ⪰ X by transitivity. This contradicts
the hypothesis that X ≻ Y ! We can then conclude that X ⪯ Z cannot be true.

Exercise. Use proof by contradiction to show that for any complete and transitive
preference,

if X ≻ Y and Y ≻ Z, then X ≻ Z

Exercise. Verify that the following three statements are logically equivalent

– U (X) ≥ U (Y ) if and only if X ⪰ Y

– U (X) > U (Y ) if X ≻ Y and U (X) = U (Y ) if X ∼ Y ;
– X ≻ Y if U (X) > U (Y ) and X ∼ Y if U (X) = U (Y ).

ˆ Since we can represent a rational preference using a utility function, we can also prove
the remark using the definition of utility function.

– This illustrates the tractability of utility maximization framework!

Example. We now show that if X ∼ Y and Y ∼ Z, then X ∼ Z. Let U (.)

represent the rational preference.

X ∼ Y and Y ∼ Z ⇔ U (X) = U (Y ) and U (Y ) = U (Z)

⇔ U (X) = U (Z)
⇔X∼Z

20
 Exercise. Show the following statements using the definition of utility function.

– If X ⪰ Y and Y ∼ Z, then X ⪰ Z.
– If X ≻ Y and Y ∼ Z, then X ≻ Z.
– If X ≻ Y and Y ⪰ Z, then X ≻ Z.

A.2 Construction of utility function

ˆ Consider a setting with M > 2 distinct options {X1 , X2 , . . . , XM }

ˆ To simplify the exposition, assume household is not indifferent between any two
options. Extension to allow indifference is straightforward but very tedious

ˆ We now describe how to construct a utility function for a given rational preference

ˆ Now imagine that we arrange M boxes in a line and label them {1, 2, . . . , M } from
left to right.

ˆ We assign the options into these boxes one-by-one in the following manner.

– Step 1: Put X1 into box 1

– Step 2: Compare X2 with X1 . If X2 ≻ X1 , then put X2 in box 1 and move X1
to box 2
..
.
– Step i: Compare Xi with existing options in boxes from left to right. If we find
Xi more preferred to the option in box j, then move the option in every box
k ≥ j to the next box on the right, and put Xi in box j instead. If we find Xi
less preferred to all options in first i − 1 box, then we put it in box i
..
.
– All the boxes are occupied by step M . X (i) denotes the option end up in box i.

ˆ The above procedure always applies because preference is complete

ˆ Now we pick any two boxes, say 2 and 5. We are going to argue that X (4) ≻ X (5) ,
X (3) ≻ X (4) and X (2) ≻ X (3) .

21
ˆ First consider X (4) and X (5) .

– Suppose the assignment of X (5) was later than that of X (4) , we must have
compared X (5) to X (4) and found that X (4) ≻ X (5) .
– If X (4) was assigned later than X (5) , we must have compared X (4) to X (5) and
found that X (4) ≻ X (5) .

ˆ Similarly, if we trace back the assignment for X (2) , X (3) and X (4) , it must be X (2) ≻
X (3) and X (3) ≻ X (4) . By transitivity, X (2) ≻ X (5) .

ˆ This argument applies to any pairs of options. Hence, the above procedure ensures
that box i contains the household’s “i-th most preferred option”

X (1) ≻ X (2) ≻ X (3) ≻ · · · ≻ X (M )

ˆ We then assign utility level M − i the option in box i. This defines a utility function.

22
A.3 Empirical evidence against rational choice
ˆ If a household’s preference is rational, she makes the same choice from the same
feasible set.

ˆ A well-known criticism of the theory of rational choice is that households’ choices

often depend on how the options are described. Specifically, a household may “perceive”
two situations differently and make different choices even though the feasible sets are
objectively identical. More importantly, such violations are often systematic, and
thus predictable.

ˆ The following examples are based on the 1983 American Psychological Association
award address, Kahneman, D., & Tversky, A. (1984). Choices, values, and frames.
American psychologist, 39(4), 341.

Example. Imagine that you are about to purchase a furniture and choose between
two means of payment, cash and credit card. Consider the following two scenarios

– The selling price is $1000. If you pay in cash instead of using credit card,
you will have a $20 discount.
– The selling price is $980. However, a $20 surcharge applies if you pay by
credit card.

Will your choice be different in the two scenarios?

In the two scenarios, the feasible sets are the same. It is basically a choice between
paying $980 in cash and paying $1000 using credit card.

Yet many people avoid the credit card surcharge but are willing to forgo the cash
discount. In particular, Thaler (1980) documented that credit card companies
in U.S. lobbied that any difference between cash and credit card customers must
take the form of a cash discount rather than a credit card surcharge.

Example. Imagine that you have decided to see a play and arrived at the theatre.
Consider the following two scenarios

23
 – You already bought a ticket for the show for $10 beforehand. However, you
discover that you have lost the ticket. The seat was not marked and the
ticket cannot be recovered. Will you pay $10 for another ticket?
– You have not bought the ticket. You are about to buy one. When you take
your wallet out, you discover that you have lost a $10 bill. You still have
enough cash to buy a ticket. Will you pay $10 for the ticket?

Will your choice be different in the two scenarios?

The feasible sets are the same in both scenarios. The first option is to purchase
the ticket seeing the play and the second option is to simply leave the theatre.

Kahneman and Tversky (1984) documented that in their questionnaire, 88% of

the subjects would buy the ticket if they lost $10 bill. However, only 46% of them
would buy a new ticket if they lost their tickets.

Exercise. Imagine that you are about to buy purchase a jacket for $125 and a
calculator for $15 in a shop. Another department store is located 20 minutes
drive away and sells the same items. Consider the following two scenarios

– Your friend tells you that the calculator is on sale for $10 in the department
store nearby. Will you visit and buy at the department store?
– Your friend tells you that the jacket is on sale for $120 in the department
store nearby. Will you visit and buy at the department store?

1. Will your choice be different in the two scenarios?

2. Describe the feasible sets in the two scenarios. Are the optimal options
different in both scenarios?

24