Lecture 7.

Substitution and Income Effects;

elasticity

A. Banerji
February 17, 2025
Demand Curve: Own Price and Demand

• In the consumer model, this translates to how a price change affects

Marshallian demand for a good.

1
Marshallian Demands

• Marshallian demands come from the bundle of goods that maximizes

utility on the budget set:
• (x1∗ (p1 , p2 , m), x2∗ (p1 , p2 , m)) = argmax u(x1 , x2 ) on B(p1 , p2 , m)
• For the record, also recall that the maximized utility as a function of
prices and income is called the indirect utility function:
• v (p1 , p2 , m) = u(x1∗ (p1 , p2 , m), x2∗ (p1 , p2 , m))

2
Demand curve

• Response of x1 to change in p1 modelled as: what is the sign of

∂x1∗ (p1 , p2 , m)/∂p1 .
• This is an own-price effect.
• Sign of ∂x2∗ (p1 , p2 , m)/∂p1 is a cross-price effect.

3
Demand curve - own price effect

• Is ∂x1∗ (p1 , p2 , m)/∂p1 < 0 so obvious?

4
Demand curve - own price effect

• Is ∂x1∗ (p1 , p2 , m)/∂p1 < 0 so obvious?

• No!

4
Demand curve - own price effect

• Is ∂x1∗ (p1 , p2 , m)/∂p1 < 0 so obvious?

• No!
• For example, if p1 decreases, and you continue consuming the
bundle you were consuming, you have income left over.
• How do you spend this leftover income? If good 1 is a normal good,
you consume more of it if your income increases; if it is an inferior
good, you consume less as your income increases.
• At the same time, you would want to substitute towards the now
relatively cheaper deal.

4
Income and Substitution Effect: Normal Good.

x2

x0∗ = h0∗

xn∗
hn∗

x1
SE1 IE1

5
Decomposing effect of price change

• How to measure that you are richer even though m is unchanged,

when p1 decreases?
• One possibility: From the new situation, remove that amount of
income which will leave you at the original level of utility, when you
were optimizing before the price change.
• This is Hicks’ ‘compensation’.
• Alternatively, we could remove the amount of income which will
permit you to just buy your pre- price change bundle of goods. This
is Slutsky compensation.

6
Hicksian Demand

• The picture was for Hicks compensation, which we focus on.

• It shows the following: p1 decreases, swivelling the budget line; so,
the price ratio p1 /p2 decreases, but you are also richer in the sense
of having income left over at the original optimum.
• At the new price ratio, reduce income (shift new budget line
inwards), until you are at a tangency with original max level of utility.
∗
• Notice that the consumption hn1 of good 1 at this intermediate
∗
point is greater than the old consumption x01 .
• This is the substitution effect: it is negative because the indifference
curve is downward sloping and strictly convex. It captures the pure
effect of the relative price change, on demand.

7
Decomposition - Substitution effect

• Mathematically, write x2 = f (x1 ) for the indifference curve: so its

slope f 0 (x1 ) < 0, and curvature f 00 (x1 ) > 0. Starting at (x01
∗ ∗
, x02 ), a
tangency when price ratio is p1 /p2 , if good 1 price decreases to
p10 < p1 , the slope p10 /p2 is less negative than p1 /p2 . The new
tangency at hn∗ on the same indifference curve happens at a slope of
f that is less negative: given f 0 < 0, f 00 > 0, this happens at a
higher x1 (and so lower x2 ).
• Economically, suppose initially MRS(x∗0 ) = p1 /p2 = 1 and
p10 /p2 = 1/2. At the new price ratio, I can sell 1 of good 1 and buy
1/2 a unit of good 2; or I can sell 1 unit of good 2 and buy 2 units
of good 1. Since MRS(x∗0 ) = 1, I am indifferent when I give up 1
unit of good 2 and get 1 of good 1; but now the market permits me
to give up 1 unit of good 2 and get 2 units of good 1; making me
better off. So, I will substitute away from good 2 into more of good
1.
8
Decomposition - income effect

• Note we say negative substitution effect because the Hicksian

demand increases as p1 decreases.
• In ‘real’ terms, income has increased. If the pure effect of this is to
increase the demand for a good, we say the good is a normal good.
∗
• In the figure, the new Marshallian demand for good 1 is xn1 . Since
∗ ∗
xn1 − hn1 > 0 happens as real income increases, the picture is that
of a normal good 1.

9
Slutsky Decomposition

• Substitution effect is always negative.

• While income effect can be positive or negative.

• For normal good

∂xi∗ ∂hi∗ ∂xi∗ ∗
= − xi .
∂pi ∂pi |∂m{z }
|{z}
− +

• For inferior goods

∂xi∗ ∂hi∗ ∂xi∗ ∗
= − xi .
∂pi ∂pi |∂m{z }
|{z}
− −

10
Slutsky Decomposition: interpretation

∂xi∗ ∂hi∗ ∂xi∗ ∗

= − xi
∂pi ∂pi |∂m{z }
|{z}
− +

Suppose there is a unit change (say a decrease) in pi .

Substitution effect: At the original level of utility, I will re-optimize my
consumption bundle to consume more of good i.
Income effect: At the original level of good i, xi∗ , I ’save’ one rupee per
unit of the good, so I save an income of xi∗ : this is my first-order saving
of income. My increase in good i consumption owing to this increase in
∂x ∗
income, is, to first order, ∂mi xi∗ .

11
Inferior Goods

• If Marshallian demand for a good decreases as income increases, we

call this an inferior good.
• Empirically, as incomes have increased with countries growing richer,
we notice consumption of some food items have fallen.
• This is different from budget share on that food item being reduced:
that has happened for the food sector more generally with increase
in incomes, and by itself has led to some structural transformation in
economies that have then allocated more resources to producing
manufacturing and services.
• This is seen in the shrinking of the share of agriculture in GDP as a
country grows richer.

12
Income and Substitution Effect: Inferior Good.

x2

xn∗
x0∗ = h0∗

hn∗

IE1

x1
SE1

13
Giffen Goods

• Does, keeping all other factors fixed, quantity demanded always

increase as own price decreases, and decrease as own price
increases?

• It is definitely true for normal goods, but may not be true for inferior
good as
∂xi∗ ∂hi∗ ∂xi∗ ∗
= − xi .
∂pi ∂pi |∂m {z }
|{z}
− −

• Inferior goods for which the income effect dominates the

substitution effect are called Giffen goods.

14
Giffen Goods:

x2

x0∗ = h0∗

hn∗

x1
SE1

15
Giffen Goods:

x2

xn∗

x0∗ = h0∗

hn∗

IE1

x1
SE1

16
Do Giffen Goods Exist in Reality?

Link

17
Jensen-Miller AER2008 on Giffen Goods

• Problem in estimating demand response: identification.

• Example: Economists claimed potatoes were a Giffen good. As
potato price increased during Irish potato famine of the 1840s, poor
Irish consumed more potatoes.
• Hard evidence of this does not exist. Potato famine itself due to
negative supply shock: supply curve moved up and left; therefore
price went up. And in the aggregate, potatoes were in short supply,
so on the aggregate potato consumption could not have increased.
• Jensen and Miller randomly assigned staple discount vouchers to
very poor consumers in China: for rice in Hunan (South) and for
wheat in Gansu (North).

18
Jensen Miller: Giffen goods

• They find, especially for rice in Hunan, that giving the subsidy on
rice (decreasing its price) decreased its consumption, and
removing the subsidy increased consumption.
• Consider caloric story: I am poor, must meet a target of 41,500
calories every 20-21 days, and earn 560 rupees.
• p1 = 40, p2 = 400 are unit prices of starch (say via rice) and protein
(dals/chicken). My utility max occurs at x1∗ = 10, x2∗ = 0.4 (in kg).
Calories are these weights in grams multiplied by 4: So,
40, 000 + 1, 600 = 41, 600.

19
Jensen Miller 2008 continued

• Now, p1 increases to 50 rupees, other prices and income is constant.

Will I consume less rice? If I consume even the earlier 10 kg, I now
spend 50 × 10 = 500 rupees on rice; I buy 60 rupees of protein, i.e.
60/400 = 0.15 kg. Total calories = 40, 000 + 600 = 40, 600.
• To meet my minimum calorie target, I need to increase rice
consumption - the way cheaper source of calories!
• This example needs that rice has no close cheap substitutes available
in my region - which can happen in rural areas with poor distribution
networks.

20
Expenditure Function: Hicksian Demand

• Instead of maximizing utility subject to a given income, we think of

its dual: the minimum expenditure we need to achieve a given level
of utility u.
• The Hicksian demands

(h1∗ (p1 , p2 , u), h2∗ (p1 , p2 , u)) = argminx1 ,x2 ≥0 p1 x1 + p2 x2 ,

such that
u(x1 , x2 ) ≥ u.
• And the minimized expenditure
e(p1 , p2 , u) = minx1 ,x2 ≥0 p1 x1 + p2 x2 s.t. u(x1 , x2 ) ≥ u is called the
expenditure function.

21
Expenditure Minimization:

• We set up the Lagrangian

L(x1 , x2 , η) = p1 x1 + p2 x2 + η(u − u(x)).

• Interior solution will satisfy first order condition for minimization

∂u(h1∗ , h2∗ )
p1 = η ∗
∂x1
∂u(h1∗ , h2∗ )
p2 = η ∗ .
∂x2

22
Expenditure Minimization:

• For interior solution we have

∂u(h1∗ ,h2∗ )
∂x1 p1
∂u(h1∗ ,h2∗ )
= .
p2
∂x2

• Marginal rate of substitution at h∗ is equal to p1 /p2 .

• and
u(h1∗ , h2∗ ) = u.
• This is the same tangency condition as for utility maximization.

23
Expenditure Minimization vs Utility Maximization:

x2

h∗

u(x) = u

x1

24
U Max - Expd Min Upshot

• We focus on strictly convex indifference curves and unique interior

optima.
• The diagrams, and the first-order conditions earlier, show that the
same tangency holds for the problems:(i)
max u(x1 , x2 ) s.t. p1 x1 + p2 x2 = m; and (ii)
min p1 x1 + p2 x2 s.t. u(x1 , x2 ) = u
• But, given the same (p1 , p2 ), when are the optima (x1∗ , x2∗ ) of (i)
and (h1∗ , h2∗ ) of (ii) equal?
• Clearly, this happens when the budget line and utility level for the 2
problems are the same: i.e. when m = e(p1 , p2 , u) and
v (p1 , p2 , m) = u.

25
Marshallian Demand and Hicksian Demand:

x2 x2

x∗ h∗

u(x) = v (p, m)
m = e(p, u)
x1 x1

26
Marshallian and Hicksian Demand

• So, if in utility maximization the income is m, then for Marshallian

and Hicksian demands to be equal, the u in expenditure
minimization needs to be v (p1 , p2 , m).
• Similarly, if the target utility in the expenditure minimization
problem is u, then the income in utility maximization is required to
be e(p1 , p2 , u).

27
Marshallian Demand and Hicksian Demand:

• Duality therefore gives us

xi∗ (p, m) = hi∗ (p, v (p, m))

and

xi∗ (p, e(p, u)) = hi∗ (p, u).

28
Own Price Effect:

• We have
hi∗ (p, u) = xi∗ (p, e(p, u)).

29
Own Price Effect:

• We have
hi∗ (p, u) = xi∗ (p, e(p, u)).

• Let’s differentiate this with respect to pi

∂hi∗ ∂x ∗ ∂x ∗ ∂e(p, u)
= i + i .
∂pi ∂pi ∂m ∂pi

29
Own Price Effect:

• Note that
∂e(p, u)
= hi∗ (p, u).
∂pi
• For, e(p1 , p2 , u) = p1 h1∗ + p2 h2∗ + η(u − u(h1∗ , h2∗ )), as the constraint
adds 0.
∂e(p,u) ∂h∗ ∂h∗ ∂h∗ ∂h∗
• ∂p1 = h1∗ + p1 ∂p11 + p2 ∂p21 − ηu1 (h1∗ , h2∗ ) ∂p11 − ηu2 (h1∗ , h2∗ ) ∂p21
∂h∗ ∂h2∗
= h1∗ + ∂p11 (p1 − ηu1 (h1∗ , h2∗ )) + ∂p1 (p2 − ηu2 (h1∗ , h2∗ ))
= h1∗
• This is an envelope result: For a small change in p1 , the minimal
expenditure changes only by the corresponding good 1 amount that
was being demanded: h1∗ ; the change in h1∗ , h2∗ get cancelled out by
the first-order conditions.

30
Own Price Effect

• This implies
∂hi∗ ∂x ∗ ∂x ∗
= i + i hi∗ (p, u).
∂pi ∂pi ∂m

31
Own Price Effect:

• Since hi∗ (p, u) = xi∗ (p, e(p, u)), we have

∂hi∗ ∂x ∗ ∂x ∗
= i + i xi∗ .
∂pi ∂pi ∂m

• Which can be written as

∂xi∗ ∂hi∗ ∂xi∗ ∗
= − xi .
∂pi ∂pi |∂m{z }
|{z}
substitution effect income effect

• This says for example: a small decrease in pi swivels Hicksian

demand along the old indifference curve at old utility level, say u. In
addition this decrease increases income by xi∗ , and per unit of that,
∂x ∗
demand changes by ∂mi

32
Own Price income and substitution effects

• To summarize, suppose p1 decreases by a tiny amount, to p10 . The

substitution effect on good 1 is the change in its demand along the
same indifference curve: ∂h1∗ /∂p1 .
• The income effect, by the Chain Rule, is the change in income,
times the change in demand for good 1 in response to change in
income: this latter term equals ∂x1∗ /∂m.
• Why is the change in income x1∗ ? Think of it as resulting from a
price decrease of a small unit Rupee 1. Then, on the x1∗ units
purchased, I save 1 rupee for each of those, totalling up to an
income of x1∗ .

33
Cross Price Effect:

the effect of other good price change on Marshallian demand.

34
Cross Price Effect:

• We have
hi∗ (p, u) = xi∗ (p, e(p, u)).

• Let’s differentiate this with respect to pj

∂hi∗ ∂x ∗ ∂x ∗ ∂e(p, u)
= i + i .
∂pj ∂pj ∂m ∂pj

35
Cross Price Effect:

• We have already shown that

∂e(p, u)
= hj∗ (p, u).
∂pj

• This implies
∂hi∗ ∂x ∗ ∂x ∗
= i + i hj∗ (p, u).
∂pj ∂pj ∂m

36
Cross Price Effect:

• Since hj∗ (p, u) = xj∗ (p, e(p, u)), we have

∂hi∗ ∂x ∗ ∂x ∗
= i + i xj∗ .
∂pj ∂pj ∂m

37
Cross Price Effect:

• Since hj∗ (p, u) = xj∗ (p, e(p, u)), we have

∂hi∗ ∂x ∗ ∂x ∗
= i + i xj∗ .
∂pj ∂pj ∂m

• Which can be written as

∂xi∗ ∂hi∗ ∂xi∗ ∗
= − xj .
∂pj ∂pj |∂m{z }
|{z}
substitution effect income effect

37
Own vs Cross Price Change:

• Own price change (hi∗ (p, u) = xi∗ (p, e(p, u)))

∂xi∗ ∂hi∗ ∂xi∗ ∗ ∂hi∗ ∂x ∗

= − xi = − i hi∗ .
∂pi ∂pi |∂m{z } ∂pi ∂m
|{z}
substitution effect income effect

• Cross price change (hj∗ (p, u) = xj∗ (p, e(p, u)))

∂xi∗ ∂hi∗ ∂xi∗ ∗ ∂hi∗ ∂x ∗

= − xj = − i hj∗ .
∂pj ∂pj |∂m{z } ∂pj ∂m
|{z}
substitution effect income effect

38
Response to other price change

• Suppose there are 2 goods that for our consumer are both normal
goods as well as substitutes: e.g. chicken and fish.
• Suppose the p1 , the price of chicken, decreases. We might expect
that the consumer will increase chicken consumption and reduce fish
consumption.
• But there is also an income effect: the consumer now has more real
income. Since fish is a normal good, the income effect on fish will
be positive.
• So, even though ∂h∗ 2/∂p1 > 0 tends to reduce fish demand, the
overall effect ∂x2∗ /∂p1 may be negative: overall, fish demand may
increase if the income effect dominates the substitution effect.

39
Cobb-Douglas Utility:

• The Marshallian demand is given by

αm βm
x1∗ = , x2∗ = .
(α + β)p1 (α + β)p2

• The Hicksian demand is given by

1
  α+β   β 1
  α+β   α
u αp2 α+β u βp1 α+β
h1∗ (p, u) = , h2∗ (p, u) = .
A βp1 A αp2

• Assume A = 1, α + β = 1

40
Cobb-Douglas Utility: A = 1, α + β = 1

• The Marshallian demand is given by

αm (1 − α)m
x1∗ = , x2∗ = .
p1 p2

• The Hicksian demand is given by

 (1−α)  α
p2 p1
h1∗ (p, u) = uαB , h2∗ (p, u) = u(1 − α)B .
p1 p2

41
Cobb-Douglas Utility: A = 1, α + β = 1

• Substitution effect
∂h1∗ ∂h2∗
= −uα(1 − α)Bp21−α p1α−2 , = uα(1 − α)Bp2−α p1α−1 .
∂p1 ∂p1

• Income effect
∂x1∗ ∂x2∗
x1∗ = uα2 Bp21−α p1α−2 , x1∗ = uα(1 − α)Bp2−α p1α−1 .
∂m ∂m

42
Own and Cross Price Effect.

• Effect of p1 on x1∗

∂x1∗ ∂h1∗ ∂x ∗
= − 1 x1∗ = −uαBp21−α p1α−2 .
∂p1 ∂p1 ∂m

• Effect of p1 on x2∗

∂x2∗ ∂h2∗ ∂x ∗
= − 1 x2∗ = 0.
∂p1 ∂p1 ∂m

43
Price Elasticity of Demand:

• In 1993 Phillip Morris announced a 18% price cut.

• Sales increased by 12.5%.

• Philip Morris’s stock fell 26% ($13.4 billion).

• Stocks of Nabisco, P & G, Coca-Cola also felt the heat.

44
Market Demand

• Consumer i demand function for commodity 1 is

x1i∗ (p, mi ).

• When all consumers are price takers, the market demand function
for commodity 1 is
n
X
x1∗ (p, m1 , . . . , mn ) = x1i∗ (p, mi ).
i=1

• Identical consumers

x1∗ (p, mn) = nx1i∗ (p, m).

45
Elasticity

• Elasticity measures the “sensitivity” of one variable with respect to

another.

• The elasticity of variable x with respect to variable y is

%4x
εx,y = .
%4y

46
Elasticity in Economics:

• Quantity demanded of commodity i with respect to the price of i

own-price elasticity of demand.

• Quantity demanded of commodity i with respect to the price of j

cross-price elasticity of demand.

• Quantity demanded for commodity i with respect to income

income elasticity of demand.

47
Own-Price Elasticity of Demand: Point Elasticity.

• Own price elasticity is defined as

4x1∗
%4x1∗ x1∗
εx1∗ ,p1 = = 4p1
.
%4p1 p1

• Why not just use the slope of a demand curve to measure the
sensitivity of demand to a change in own price?

• Because the value of sensitivity then depends upon the units of

measurement used for quantity demanded.

48
Elasticity:

• Own price elasticity demand is symbolized by ε

1. ε < −1 , demand is elastic

2. ε > −1, demand is inelastic

3. ε = −1, demand is unitary elastic

4. ε = 0, demand is completely inelastic

5. ε = −∞, demand is infinitely elastic

49
Example:

• Consider the linear demand

a − p1 p1
x1 = , ε=− .
b a − p1

p1

ε = −∞

∞ < ε < −1

a
2 ε = −1

−1 < ε < 0
x1
a ε=0
2b

50
Cross-Price Elasticity of Demand:

• Cross price elasticity is defined as

4x1∗
%4x1∗ x1∗
εx1∗ ,p2 = = 4p2
.
%4p2 p2

51
Income Elasticity of Demand:

• Income elasticity is defined as

4x1∗
%4x1∗ x1∗
εx1∗ ,m = = 4m
.
%4m m

52