Lecture 6: Welfare

University of Queensland
So far . . .

I How changes in income and prices affect consumed quantity

I Price Effect = Substitution Effect + Income Effect

I Slutsky Equation
Welfare

I So far we have studied changes in quantities

I But sometimes we wonder how the well-being of a consumer

is affected by a price change

I You may think that we can look at how the consumer’s utility
changes, but

I Utility numbers have no meaning

I Utility numbers cannot be aggregated across people

Two Ways: Compensating Variation
and Equivalent Variation
Two Ways of Translating Welfare into Money

Compensating Variation (CV)

The amount of money we would need to give to (or take away
from) the consumer under the new prices so that s/he is just as
well off as before
I the amount that we need to compensate the consumer for
undergoing the price change

Equivalent Variation (EV)

The amount of money that the consumer is willing to pay (or
must be paid) under the old prices so that s/he is just as well off
as after the price change
I the amount that can be taken away from (or give to) the
consumer for avoiding the price change
Compensating Variation (Price Increase)

y
MC

CV

M0

B A

U0
U1
x
Compensating Variation and Expenditure Minimisation

I Suppose Px changes to Px0

I If we run expenditure minimisation using the old utility level

(call it U0 ) with the new prices, we obtain the expenditure
required to achieve the old utility level under the new prices,
E ∗ (Px0 , Py , U0 )

I Compensating variation is the difference between

E ∗ (Px0 , Py , U0 ) and the original income

CV = E ∗ (Px0 , Py , U0 ) − M = E ∗ (Px0 , Py , U0 ) − E ∗ (Px , Py , U0 )

CV is the area left of the original Hicksian Demand

y
MC
CV
M0
H
B A
U1 U0
x
Px
D h (U0 )
P1
P0 Dm
x
xB xH xA
CV is the area left of the original Hicksian Demand II

CV = E ∗ (Px0 , Py , U0 ) − E ∗ (Px , Py , U0 )
Z Px0
∂E ∗ (px , Py , U0 )
= dpx
Px ∂px
Z Px0
= x H (px , Py , U0 )dpx
Px

where the last equality follows from looking at the Lagrangian of

the Expenditure minimisation problem evaluated at the solution
(the Hicksian demands):

min M = px x H + Py y H + µ(U − U(x H , y H ))

x,y ,µ

= E ∗ (px , Py , U0 )

. . . but don’t worry about the technical details in this derivation

Equivalent Variation (Price Increase)

M0
EV
ME

B A
H
U0
U1
x
Equivalent Variation and Expenditure Minimisation

I Suppose Px changes to Px0

I If we run expenditure minimisation using the new utility level

(call it U1 ) with the old prices, we can obtain the minimal
expenditure required to achieve the new utility level under the
old prices, E ∗ (Px , Py , U1 )

I Equivalent variation is the difference between E ∗ (Px , Py , U1 )

and the original income

EV = E ∗ (Px , Py , U1 ) − M = E ∗ (Px , Py , U1 ) − E ∗ (Px0 , Py , U1 )

EV is the area left of the new Hicksian Demand

M0
EV
ME
B A
H U0
U1
x
Px

P1
P0 Dm
D h (U1 ) x
xB xH xA
Change in Consumer’s Surplus
Area left of the Marshallian Demand is called the change in
Consumer’s Surplus (when the price of x changes)

M0

B A
U1 U0
x
Px

P1
P0 Dm
x
xB xA
CV, EV and Change in Consumer’s Surplus
Normal Good, Price Increase

Px
EV = S
D h (U0 ) CS = S + T
P1 CV = S + T + U
U
S
T
P0
Dm
D h (U1 )

Qx
xB xA
CV, EV and Consumer’s Surplus
I Depending on

1. Whether the good is normal or inferior; and

2. Whether it is a price increase or decrease

the picture of CV, EV and Consumer’s Surplus may look

somewhat different

I Yet the change in Consumer’s Surplus is always between CV

and EV

I Thus Consumer’s Surplus change is a handy approximation of

the monetary measure of welfare change

I If the Hicksian and Marshallian demands are the same,

Consumer’s Surplus change will be an exact measure
Question: When will they be the same?
Example: Staff Discounts
Featured Example: Staff Discounts

I Some retailers (e.g., apparels, cosmetics) offer their staff to

buy their products at a significant discount

I Is this a profitable practice for retailers?

I How about eliminating the staff discount, and use the cost of
the staff discount scheme to fund a pay raise?
Staff Discount: setup

I Suppose a staff member consumes only two goods: the

product of the retailer (good x) and money for all other goods
(good y )

I Normalise Py to 1

I Let Px be the recommended retail price for x

I Suppose staff member pays only δPx for each unit of x, with

0<δ<1

I Let M be the staff salary without discount

Staff Discount: Indifference Curves

y
M

U0
x
x0 M
Px
Staff Discount: Benefits to the Retailer

I When the staff member receives a staff discount on x, she can

be paid less and still receive the same utility as before

I In particular, she requires only E ∗ (δPx , 1, U0 ) to maintain the

old utility level U0

I Thus with the staff discount, the retailer can save

M − E ∗ (δPx , 1, U0 ) = E ∗ (Px , 1, U0 ) − E ∗ (δPx , 1, U0 )

in salary payment

I This is the (negative of) the compensating variation associated

with the staff discount
Staff Discount: Cost to the Retailer

Px

Px

δPx
x H (Px ; 1, U0 )
x
x0 x1
Staff Discount: Overall

I So essentially it is more cost-effective to “buy” utility for the

staff with a pay raise than a staff discount

I This should come at no surprise — subsidy in cash is better

than subsidy in kind in terms of welfare

I However, do note that the net cost of staff discount depends

on the slope of the Hicksian Demand curve: the steeper it is,
the smaller is the net cost
Staff Discounts: Why do they exist?

I Many claim that it is a marketing strategy — staff members

essentially become models for the products

I Tax and Labour laws — the cost of a $1 pay raise is often

larger than $1 (super contribution, leave loading, etc.)

I Staff purchases may take up idle capacity (e.g., Staff family

airfares, Uni Staff discounts)
Revealed Preference Approach –
Non-compulsory/ non-assessable
Revealed Preference Approach

I So far we have assumed utility-maximisation, and discussed its

implications on choices

I We can take another angle: What kind of choices will be

consistent with rational decision making?

I This approach is known as revealed preference

Revealed Preference

I Suppose there are two distinct bundles, A and B

I We say A is directly revealed preferred to B (also written as

A R B) if:
When A is chosen, B is affordable

I In other words, A is chosen over B

Revealed Preference: Example

I Suppose Px = 2, Py = 3 and
the consumer’s income is 24

I Bundle A is chosen y

Bundle Cost
A = (6, 4) B
B = (3, 6) C
C = (5, 5) A
D = (7, 2)
D
x
I A is revealed preferred to B and
D
I A is not revealed preferred to C
Weak Axiom of Revealed Preference (WARP)

I The Weak Axiom of Revealed Preference (WARP) says that,

For any two distinct bundles A and B
If A is directly revealed preferred to B
Then B cannot be directly revealed preferred to A

I In plain language: if A is chosen over B, then B cannot be

chosen over A
An Example in which WARP is satisfied
y

B
x

I When A is chosen, B is affordable (A R B)

I When B is chosen, A is not affordable (B 6R A)

Another Example in which WARP is satisfied
y

B
A
x

I When A is chosen, B is not affordable (A 6R B)

I When B is chosen, A is affordable (B R A)

Yet Another Example in which WARP is satisfied
y

I When A is chosen, B is not affordable (A 6R B)

I When B is chosen, A is not affordable (B 6R A)

An Example in which WARP is violated
y

A
x

I When A is chosen, B is affordable (A R B)

I When B is chosen, A is affordable (B R A)

WARP as a Pairwise Consistency Condition

I WARP ensures that pairwise choices are consistent

I However, since it is only a pairwise condition, it does not

guarantee that comparisons between three or more bundles
make sense
A Three-Good Example

Prices Chosen Bundle Cost

Px Py Pz Name x y z Bundle A Bundle B Bundle C
2 3 3 A 3 1 7
6 4 6 B 7 3 1
9 9 6 C 1 7 3

I Is A revealed preferred to B? I Is B revealed preferred to A?

I Is B revealed preferred to C ? I Is C revealed preferred to B?

I Is C revealed preferred to A? I Is A revealed preferred to C ?

Three-Good Example: Explanations
I In this three-good example, WARP is satisfied
I A is revealed preferred to B and B is not revealed preferred to
A
I B is revealed preferred to C and C is not revealed preferred to
B
I C is revealed preferred to A and A is not revealed preferred to
C

I However, notice that the “revealed preferred to” relation is

not transitive:
I A is revealed preferred to B
I B is revealed preferred to C
I But C is revealed preferred to A!

I To fix this problem, we need a concept stronger than WARP

Indirect Revealed Preference

I Consider two bundles, A and B

I A is indirectly revealed preferred to B if there is a chain of

bundles, C1 , C2 , . . . , Cn , such that

1. A is directly revealed preferred to C1 ;

2. For each i = 1, 2, . . . n, Ci is directly revealed preferred to

Ci+1 ; and

3. Cn is directly revealed preferred to B.

Strong Axiom of Revealed Preference (SARP)

I The Strong Axiom of Revealed Preference (SARP) says that,

For any two distinct bundles A and B,

If A is directly or indirectly revealed preferred to B
Then B cannot be directly or indirectly revealed preferred
to A
Importance of SARP

Theorem
A consumer’s set of observed choices can be generated by the
maximisation of a (well-behaved) utility function if and only if it
satisfies SARP.

I In other words, SARP gives us ALL the restrictions utility

maximisation places on observed choices.
Uses of Revealed Preference

I Revealed preference treats the process of making choices as a

black box

I This makes it quite useful in applications when we are

skeptical of writing down a specific (form of a) utility function

I Here is an academic research article that uses only revealed

preference to determined the impact of a welfare reform in the
US:
Kline, Patrick and Melissa Tartari. “Bounding the Labor
Supply Responses to a Randomized Welfare Experiment: A
Revealed Preference Approach.” American Economic Review,
106(4): 972–1014
Trapping the Indifference Curves

I Utility, hence indifference curves, are not observable

I However, SARP, together with mild assumption on preferences

(specifically, monotonicity and convexity), allows us to “trap”
the position of indifference curves with limited information

I This can be quite useful in policy evaluations

Trapping the Indifference Curve from Choices

Key:
C
Strictly Worse than A

Strictly Prefers to A

D
A B
x
Your Notes on Trapping the Indifference Curve
Summary

I Measuring the change in welfare

I Compensated Variation and Equivalent Variation

I Application: Staff discounts