Lecture 3 - Axioms of Consumer Preference and the Theory of

Choice
David Autor, MIT Economics and NBER

14.03/14.003 Microeconomic Theory and Public Policy, Fall 2016

1
A brief outline
Topics for this Lecture Note
1. Consumer preference theory

(a) Notion of utility function

(b) Axioms of consumer preference
(c) Monotone transformations

Where we are headed

Theory

1. Consumer preference theory

2. Theory of choice

(a) Constrained utility maximization

(b) Constrained expenditure minimization

3. Individual demand functions

• Consumer demand holding the consumer’s budget constant

• Consumer demand holding the consumer’s utility constant
• Linking these two demand functions

4. Market demand

• Market demand for a single good as a function of its price (partial equilibrium)
• Market demand for all goods as a function of their relative prices (general equilibrium)

Applications

1. The Carte Blanche principle and the Dead Weight Loss of Christmas

2. What’s the best way to aid the poor: anti-poverty programs or cash?

3. Do wage subsidies affect employment?

4. Giffen goods: Do they exist, and are they relevant?

2
1 Consumer Preference Theory

A consumer’s utility from consumption of a given bundle “A” is determined by a personal utility
function.

1.1 Cardinal and ordinal utility

• Cardinal utility function
According to this approach U (A) is a cardinal number, that is:
U : consumption bundle −→ R1 measured in “utils”

• Ordinal utility function

More general than cardinal utility function
U provides a “ranking” or “preference ordering” over bundles.


P
 A B

U : (A, B) −→ BP A

AI B


Used in demand/consumer theory

Cardinal vs Ordinal Utility Functions

The problem with cardinal utility functions comes from the difficulty in finding the appropriate
measurement index (metric).

• Is 1 util for person 1 equivalent to 1 util for person 2? Or if we increase a person’s utility from
1 to 2, is she twice as happy?

• By being unit-free ordinal (but not cardinal) utility functions avoid these problems.

• What’s important about a utility function is that it allows us to model how people make
personal choices, that is, how they choose among competing alternatives. We do not need to
know how many “utils” people experience from each choice to obtain this information; we just
need to know how they rank choices.

• It’s much harder—and not always meaningful, but not always meaningless—to make/model
interpersonal comparisons of utility than to study intrapersonal choices, that is, the choices
that an individual selects among different options (often, for simplicity, different bundles of
goods). It turns out that we don’t need to make interpersonal comparisons of utility to make
reach positive and even some strong normative conclusions. This is part of the power of
economic theory: it goes an amazingly long distance on a very small (but not weak) set of
assumptions.

3
1.2 The Axioms of Consumer Preference Theory
Next up on our agenda: the axioms of consumer theory. These axioms were developed for three
purposes:

1. To allow for mathematical representation of utility functions

2. To portray rational behavior (rational in this case means ‘optimizing’)

3. To derive “well-behaved” demand curves

When we get through the entire corpus of consumer theory, you may be surprised to realize that all
of the results and predictions we will have derived rest entirely, and with no further assumptions,
on the small set of axioms below.

1.2.1 Axiom 1: Preferences are Complete (“Completeness”)

For any two bundles A and B, a consumer can establish a preference ordering. That is, for any
comparison of bundles, she will choose one and only one of the following:

1. A P B

2. B P A

3. A I B

Without this property, preferences are undefined.

1.2.2 Axiom 2: Preferences are Transitive (“Transitivity”)

• For any consumer if A P B and B P C then it must be that A P C.

• This axiom says that consumers are consistent in their preferences.

1.2.3 Axiom 3: Preferences are Continuous (“Continuity”)

• If A P B and C lies within an ε radius of B then A P C.

• We need continuity to derive well-behaved demand curves.

Given Axioms 1- 3 are obeyed we can always define a utility function.

Indifference Curves

• Define a level of utility say U (x) = U . Then, the indifference curve for U , IC(U ) is the locus
of consumption bundles that generate utility level U for utility function U (x).

4
• An Indifference Curve Map is a sequence of indifference curves defined over every possible
bundle and every utility level: {IC(0), IC(ε), IC(2ε), ...} with ε = epsilon

• Indifference curves are level sets of this utility function. To see this in the graph above, imagine
you are looking onto the good x- good y plane from above and the curves down onto it. These
curves are analogous to contours on a topological map, which show all points with an equal
altitude, except here they show points with the same utility.


IC3 −→ Utility level U3 

• IC2 −→ Utility level U2 U3 > U2 > U1

IC1 −→ Utility level U1


5
 • This is called an Indifference Curve Map. Properties:

– Every consumption bundle lies on some indifference curve (by the completeness axiom)
– Indifference curves are smooth due to the continuity axiom. We don’t have sharp discon-
tinuous preference reversals.
– Any utility function that satisfies Axioms 1- 3 cannot have indifference curves that cross.

Axioms 4 and 5 are introduced to reflect observed behavior

• They simplify problems greatly, but they are not necessary for a theory of rational choice.

1.2.4 Axiom 4: Non-Satiation (Never Get Enough)

• Given two bundles, A and B, composed of two goods, X and Y , where XA = amount of X
in A, similarly XB , and YA = amount of Y in A, similarly YB . If XA = XB and YA > YB
(assuming utility is increasing in both arguments) then A P B, regardless of the levels of
XA , XB , YA , YB

• This implies that:

1. The consumer always places positive value on more consumption (if a good is a bad, e.g.,
pollution, just reverse the sign so the gad is a good, e.g., less pollution).
2. Indifference curve map stretches out endlessly (there is no upper limit to utility)

• While this axiom is not necessary to construct a well-defined utility function, it is helpful.

• Note that Banerjee calls this property “monotonicity” referring to the idea that more of a
desirable commodity cannot make a consumer worse off. While I generally reserve the word
monotonicity for our subsequent discussion of monotone transformations of utility functions,
it’s legitimate to use the phrase monotonicity of preferences to refer to the axiom of non-
satiation.

1.2.5 Axiom 5: Diminishing Marginal Rate of Substitution

• This axiom is also unnecessary to construct a well-defined utility function, but we believe it
generally captures a fundamental feature of human preferences.

• This axiom also makes the mathematics of consumer theory much simpler; it means we will
stumble across relatively few corner solutions where the consumer maximizes utility by spend-
ing the entire budget on one good.

• In order to define this axiom we need to introduce the concept of Marginal Rate of Substitution
and some further preliminary explanations.

6
• Definition: MRS measures willingness to trade one bundle for another. Example:
Bundle A = (6 hours of sleep, 50 points on the problem set)
Bundle B = (5 hours of sleep, 60 points on the problem set)
If A and B lie on the same indifference curve
A student is willing to give up 1 more hour of sleep for 10 more points on the problem set.
Her willingness to substitute sleep for grade points at the margin (i.e. for 1 fewer hours of
sleep) is:
10
−1 = −10

M RS (sleep for points) = |−10| = 10

• MRS is measured along an indifference curve and it may vary along the same indifference
curve (e.g. you may be willing to give up 20 points to get just 1 hour of sleep but are willing
to put in an extra hour for 5 points if it is the difference between 8 and 9 hours of sleep). If
so, we must define the MRS relative to some bundle (starting point).

• By definition, utility is constant along an indifference curve: dU = 0. Therefore:

∂U ∂U
0 = dx + dy
∂x ∂y
0 = M Ux dx + M Uy dy
dy M Ux
− = = MRS of x for y
dx M Uy

• Note the slightly confusing terminology. MRS of x for y is the marginal utility of x divided
by the marginal utility of y (holding total utility constant), which is equal to −dy/dx.

• MRS must always be evaluated at some particular point (consumption bundle) on the indiffer-
ence curve. So one should really write MRS(x0 , y 0 ) where (x0 , y 0 ) is a particular consumption
bundle.

• We are ready to explain what is meant by Diminishing Marginal Rate of Substitution.

7
• MRS of x for y decreases as we go down the indifference curve.

• This indifference curve exhibits diminishing MRS: the rate at which (at the margin) a consumer
is willing to trade x for y – that is, the amount of y she demands in exchange for one unit of
x – diminishes as the level of x consumed goes up.

• That is, the slope of the indifference curve between points B and C is less than the slope of
the curve between points A and B. Mathematically, this means that MRS of x for y (the ratio
M Ux /M Uy ) is decreasing in x.

• Diminishing MRS implies that consumers prefer diversity in consumption.

8
• A formal definition of diminishing MRS can be given through the mathematical notion of
convexity. (Banerjee denotes Axiom 5 as Convexity rather than Diminishing MRS. As the
following makes clear, these terms are synonymous.)

• Definition. A function U (x, y) is convex if for any arguments (x1 , y1 ) and (x2 , y2 ) where
(x1 , y1 ) 6= (x2 , y2 ):

U (αx1 + (1 − α)x2 , αy1 + (1 − α)y2 ) ≥ αU (x1 , y1 ) + (1 − α)U (x2 , y2 ),

where α ∈ (0, 1).

• The figure below shows this graphically. Suppose the two bundles, (x1 , y1 ) and (x2 , y2 ) are on
the same indifference curve. The diminishing MRS property states that the convex combination
of these two bundles is on a higher indifference curve than either of the two initial bundles.

• where x∗ = αx1 + (1 − α)x2 and y ∗ = αy1 + (1 − α)y2 . This is verified for every α ∈ (0, 1).

• A utility function U (·) exhibits diminishing MRS iff the indifference curves of U (·) are convex.

• The following is an example of a non-convex curve:

9
• In this graph not every point on the line connecting two points above the curve is also above
the curve, therefore the curve is not convex.

• Q: Suppose you love drinking coffee and eating sushi, but you strongly dislike consuming them
together. Thus, the more you consume of one (at a sitting), the less you want of the other.
How do we draw these preferences?

• [Note that we assume here that both goods are consumed at once. If you could preserve the
(fresh) coffee or the (fresh) sushi and then consume it later, you would likely have convex
preferences over sushi and coffee.]

10
 • If your indifference curves were concave as above, and holding the budget set constant, you
should not diversify consumption.

Non-crossing of indifference curves

• Using the transitivity and non-satiation axioms, we can demonstrate that indifference curves
cannot intersect.

• Proof: say two indifference curves intersect:

• According to these indifference curves:

AP B
B I C
C P D
D I A

• By the above mentioned axioms, A P D and A I D, which is a contradiction.

1.3 Cardinal vs Ordinal Utility

• A utility function of the form U (x, y) = f (x, y) is cardinal in the sense that it reads off “utils”
as a function of consumption.

• But we don’t know what utils are or how to measure them. Nor do we assume that 10 utils is
twice as good as 5 utils. That would be a cardinal assumption.

• Our notion of consumer preferences focuses on the consumer’s ranking (or ordering) over
bundles of goods, rather than the consumer’s “level” of utility for a given bundle. That is, we
want to know if A P B but not by how much.

11
• We do, however, care that the MRS along an indifference curve is well defined, i.e. we do want
to know precisely how people trade off among goods in indifferent (equally preferred) bundles.

• In consumer theory, we choose to use ordinal not cardinal utility functions. As we show below,
ordinal utility functions preserve the MRS property on which we are focused but dispense with
cardinal properties (e.g., counting utils).

• Q: How do we preserve properties of utility that we care about and believe in (a. ordering
is unique; and b. MRS exists) without imposing cardinal properties? We can weaken our
definition of utility functions by stating that a utility function is only defined up to a “mono-
tonic transformation.” That is, if utility function g 0 () is a monotone transformation of utility
function g (), we will say they are identical for purposes of consumer theory.

• Definition: Monotonic Transformation

Let I be an interval on the real line (R1 ) then: g : I −→ R1 is a monotonic transformation if
g is a strictly increasing function on I.
If g(x) is differentiable then g 0 (x) > 0 ∀x
Informally: A monotone transformation of a variable is a rank-preserving transformation.
[Note: not all rank-preserving transformations are differentiable.]

• Examples of monotone and non-monotone transformations Let y be defined on R1 :

1. x = y + 1 [monotone]
2. x = 2y [monotone]
3. x = exp(y) [monotone]
4. x = abs(y) [non-monotone]
5. x = y 2 [monotone if y ≥ 0]
6. x = ln(y) [monotone if y > 0]
7. x = y 3 [monotone if y ≥ 0]
8. x = − y1 [monotone if y ≥ 0]
9. x = max(y 2 , y 3 ) [monotone if y ≥ 0]
10. x = 2y − y 2 [non-monotone]

• If U2 (.) is a monotone transformation of U1 (.), i.e. U2 (.) = f (U1 (.)) where f (.) is monotone in
U1 as defined earlier, then:

1. U1 and U2 exhibit identical preference rankings

2. MRS of U1 (U ) and U2 (U ) are the same
3. U1 and U2 are equivalent for consumer theory

12
• Example: U (x, y) = xα y β (Cobb-Douglas)

• What is the MRS along an indifference curve U0 ?

U0 = xα0 y0β
dU0 = αxα−1
0 y0β dx + βxα0 y0β −1 dy
αxα−1 y0β

dy  0 α y0 ∂U/∂x
= − = =
dx U =U0 βxα0 y0β −1 β x0 ∂U/∂y

• Consider now a monotonic transformation of U :

U 1 (x, y) = xα y β
U 2 (x, y) = ln(U 1 (x, y))
U 2 = α ln x + β ln y

• What is the MRS of U 2 along an indifference curve?

U02 = ln U0 = α ln x0 + β ln y0
α β
dU02 = dx + dy = 0
x0 y0

dy 
 α y0 ∂U/∂x
= =
dx U 2 =U 2 β x0 ∂U/∂y
0

which is the same as we derived for U 1 .

13
Optional: General demonstration of this result

• How do we know that monotonic transformations always preserve the MRS of a utility func-
tion?

• Let U = f (x, y) be a utility function

• Let g(U ) be a monotonic transformation of U = f (x, y) and differentiable

• The MRS of g(U ) along an indifference curve where U0 = f (x0 , y0 ) and g(U0 ) = g(f (x0 , y0 ))

• By totally differentiating this equality we can obtain the MRS.

dg(U0 ) = g 0 (f (x0 , y0 ))fx (x0 , y0 )dx + g 0 (f (x0 , y0 ))fy (x0 , y0 )dy

g 0 (f (x0 , y0 ))fx (x0 , y0 )

dy  fx (x0 , y0 ) ∂U/∂x
−  = 0
= =
dx g(U )=g(U0 ) g (f (x0 , y0 ))fy (x0 , y0 ) fy (x0 , y0 ) ∂U/∂y

which is the MRS of the original function U (x, y) .

14
MIT OpenCourseWare
https://ocw.mit.edu

14.03 / 14.003 Microeconomic Theory and Public Policy

Fall 2016

For information about citing these materials or our Terms of Use, visit: https://ocw.mit.edu/terms.