Consumer Theory: The Mathematical Core Dan McFadden, C103

Suppose an individual has a utility function U(x) which is a function of non-negative

commodity vectors x = (x1,x2,...,xN), and seeks to maximize this utility function subject to the budget
constraint p@x = p1x1 + p2x2 + ... + pNxN # y, where y is income and p = (p1,p2,...,pN) is the vector of
commodity prices. Define u = V(p,y) to be the value of utility attained by solving this problem; V
is called the indirect utility function. Let x = X(p,y) = (X1(p,y),X2(p,y),...,XN(p,y)) be the commodity
vector that achieves the utility maximum subject to this budget constraint. The function X(p,y) is
called this consumer’s market or Marshallian demand function. Since this function maximizes
utility subject to the budget constraint, V(p,y) / U(X(p,y)) / U(X1(p,y),X2(p,y),...,XN(p,y)). The
figure below shows for the case of two goods the budget line d-e, and the point a that maximizes
utility.1 Note that the point a is on the highest utility contour (or indifference curve) that touches the
budget line, and that at a the indifference curve is tangent to the budget line, so that its slope, the
marginal rate of substitution (MRS) is equal to the slope of the budget line, -p1/p2.

Note that the indirect utility function 20

and the market demand functions all depend
good 2

on income and on the price vector. For

15
example, the demand for good 1, written out,
is x1 = X1(p,y). When demand for good 1 is
graphed against its own price p1, then 10
changes in own price p1 correspond to
d
movement along the graphed demand curve,
5
while changes in income y or in the cross- a
price p2 correspond to shifts in the graphed
demand curve. If all income is spent, then e
0
p@X(p,y) / y. The consumer is said to be 0 5 10 15 20 25
locally non-satiated if in each neighborhood good 1
of a commodity vector there is always
another that has strictly higher utility. When
the consumer is locally non-satiated, no commodity vector that costs less than y can give the
maximum obtainable utility level, the utility-maximizing consumer will spend all income,, and the
indirect utility function V(p,y) is strictly increasing in y. Local non-satiation rules out “fat”
indifference curves. A sufficient condition for local non-satiation is that the utility function be
continuous and monotone increasing; i.e., if xN>> xO, then U(xN) > U(xO).

1
The graph was created using the utility function U(x) = x11/3x22/3. The budget line d-e is given by 9x1 + 18x2
= 135. The demand functions that maximize utility are X1(p1,p2,y) = y/3p1 and X2(p1,p2,y) = 2y/3p2. The indirect
utility function is V(p1,p2,y) = y(3p1)-1/3(3p2/2)-2/3.

1
 If income and all prices are scaled by a constant, the budget set and the maximum attainable
utility remain the same. This implies that V(p,y) and X(p,y) are unchanged by rescaling of income
and prices. Functions with this property are said to be homogeneous of degree zero in income and
prices.

Given utility level u, let y = M(p,u) be the minimum income needed to buy a commodity
vector that gives utility u; M is called the expenditure function. Let x = H(p,u) /
(H1(p,u),H2(p,u),...,HN(p,u)) be the commodity vector that achieves the minimum expenditure subject
to the constraint that a utility level u be attained. Then,

M(p,u) / p"H(p,u) / p1H1(p,u) + ... + pNHN(p,u).

The demand functions H(p,u) are called the Hicksian, constant utility, or compensated demand
functions. If all prices are rescaled by a constant, then the commodity vector solving the minimum
income problem is unchanged, so that H(p,u) is homogeneous of degree zero in p. The expenditure
is scaled up by the same constant as the prices, and is said to be linear homogeneous in p. Suppose
in the figure above, instead of fixing the budget line and locating the point a where the highest
indifference curve touches this line, we had fixed the indifference curve and varied the level of
income for given prices to find the minimum income necessary to touch this indifference curve.
Provided the consumer is locally non-satiated so that all income is spent, this minimum again occurs
at a. Then, solving the utility maximization problem for a fixed y and attaining a maximum utility
level u, and solving the expenditure minimization problem for this u and attaining the income y, lead
to the same point a. In this case, if the consumer utility level is u = V(p,y), then y is the minimum
income at which utility level u can be achieved. Then, the result that the utility maximization
problem and the expenditure minimization problem pick out the same point a implies that y /
M(p,V(p,y)) and u / V(p,M(p,u)); i.e., V and M are inverses of each other for fixed p. Further,
X(p,y) / H(p,V(p,y)) and X(p,M(p,u)) / H(p,u).

The demand for a commodity is said to be normal if demand does not fall when income rises,
and inferior or regressive if demand falls when income rises. A commodity is a luxury good if its
budget share rises when income rises, and is otherwise a necessary good. The budget share of the
first commodity is s1 = p1X1(p,y)/y. Define the income elasticity of demand,

0 = (y/X1(p,y))@MX1(p,y)/My.

This is the percentage by which demand for good 1 increases when income goes up by one percent.
Show as an exercise that the income elasticity of the budget share satisfies

(y/ s1)@Ms1/My= 0 - 1.

2
Then, a luxury good has an income elasticity of its budget share greater than zero, a necessary good
has an income elasticity of the budget share less than zero, and an inferior good has an income
elasticity of the budget share less than -1. A graph of the demand for a good against income is called
an Engle curve. The figure below shows the Engle curves for three cases.
Engle Curves
Budget Share of Good 1 0.3 The income elasticities
Regressive of demand at I = 250 are
-0.33 for Regressive,
0.8 for Necessary,
0.2 1.2 for Luxury.

Necessary
Luxury
0.1

0
0 100 200 300
Income

It is possible to trace out the locus of demand points in an indifference curve map as income
changes with prices fixed; this locus is called an income-offer curve or income-expansion path.
Points on an income-expansion path correspond to points on Engle curves for each of the
commodities. The figure below shows an income-expansion path when good 1 is a luxury good.

Locus of Demand Points

Income-Expansion
Good 2

4
Path
3

0
0 0.5 1 1.5 2 2.5 3
Good 1

3
 In an indifference curve map, it is also possible to trace out the locus of demand points as the
price of a good changes; this locus is called an price offer curve. The figure below depicts a typical
price offer curve. Note that the price-offer curve is the locus of tangencies between indifference
curves and budget lines that pivot about one point on the vertical axis, in this case (0,4). As one
moves out along the offer curve, one is identifying quantities demanded of good 1 as its price falls.
In this diagram, falling price always results in increased demand for good 1. This is called the “law
of demand”, but a qualification is needed to make this law hold, as described below.

Locus of Demand Points

Offer Curve
Good 2

0
0 0.5 1 1.5 2 2.5 3
Good 1

If one plots the demand for good 1 against its own price, one obtains the standard picture of a
demand function, given in the next figure. Remember that this is the locus of points satisfying x1
= X1(p,y) as p1 changes, with y and p2,...,pN kept fixed. If y or p2,...,pN were to change, this would
appear as a shift in the demand function given in the picture.

It is often useful to characterize demand Demand Function

functions in terms of their own-price elasticity, 4
defined as
Price of Good 1

, = (p1/X1(p))@MX1(p)/Mp1. 2

In this definition, , is negative in the usual case

that the demand function slopes downward. 0
When , < -1, demand is said to be elastic, and 0 0.5 1 1.5 2 2.5 3
when -1 < , < 0, demand is said to be inelastic. Quantity of Good 1
Verify as an exercise that the own-price elasticity
of the budget share of good 1 is 1 + ,.

4
 If one looks in more detail in an indifference curve map at the change in demand for a good
resulting from a decrease in its price, one can decompose the total change into the effects of a pure
income change, with relative prices fixed, and a pure price change, with income compensated to
keep utility constant. The first of these effects is termed an income effect, and the second is termed
a substitution effect. What we will show is that the substitution effect always operates to increase
the demand for a good whose price falls. We will show that for a normal good, the income effect
reinforces the substitution effect, so that falling price increases demand. However, for an inferior
good, we will show that the substitution and income effects work in opposite directions. It is even
possible for the income effect to be larger in magnitude than the income effect, so that a fall in price
reduces demand. In these circumstances, the good is called a Giffen good. The occurrence of
Giffen goods is primarily a curiosity rather than a common event, and requires that a good have a
large share of the budget and be strongly inferior. An example of a Giffen good is day-old bread,
which may absorb most of the income of a homeless person. If the price of day-old bread falls, this
consumer needs to commit less income to subsistence on day-old bread, and may instead buy more
fresh bread. The figure below shows the decomposition of a total price effect into income and
substitution effects. In this figure, the good is normal, so that the income and substitution effects
are reinforcing. In the diagram, start from the budget line d-e, for which utility is maximized at a.
Now decrease the price of good 1 so the budget line becomes d-f. On this new budget line, utility
is maximized at c. The total price effect is a-c. In this case, the total own price effect is to increase
demand for good 1 (from a to c) when the price of good 1 falls. Now introduce the budget g-h
which keeps prices the same as the
were at d-e, but compensates income 20
so that maximum utility (attained at b)
good 2

is the same as it is at c. Then, a-b is

15
the income effect, and b-c is the
substitution effect. The substitution g
effect always operates in the direction 10
of increasing demand for the good d
b
whose price is decreasing. The income c
5
effect also increases demand for good a
f
1 in this case, since it is a normal
good. However, if good 1 had been an e
0
h
inferior good, the income effect would 0 5 10 15 20 25
partially (or completely) offset the good 1
substitution effect.

5
 The statement we obtained geometrically that the substitution effect always increases the
demand for a good whose price is falling can also be established mathematically, with great
generality. The remainder of these notes provide a simple mathematical exposition of this analysis.

From the definition of the expenditure function, for any vector q, and g a small scalar,

M(p+gq,u) / (p+gq)"H(p+gq,u) # (p+gq)"H(p,u) = M(p,u) + gq"H(p,u),

or
M(p+gq,u) - M(p,u) # gq"H(p,u).

This implies

limg`0(M(p+gq,u) - M(p,u))/g # q"H(p,u) # limg_0(M(p+gq,u) - M(p,u))/g.

But the left and right hand ends of this inequality both converge to q"LpM(p,u) when the derivative
exists, establishing that q"H(p,u) = q"LpM(p,u). This is true for any q, so H(p,u) / LpM(p,u). This
is called Shephard’s identity. From this, X(p,y) / H(p,V(p,y)) / LpM(p,u) . u=V(p,y).

From the definition of the expenditure function,

M(p+q.u) = (p+q)"H(p+q.u) = p"H(p+q,u) + q"H(p+q,u) $ M(p,u) + M(q,u).

A linear homogeneous function that satisfies this inequality is said to be concave. Mathematical
implications of concavity are that M will be continuous and will almost always have first and
second derivatives, the mixed second partial derivatives will be independent of the order of
differentiation, and the own second partial derivatives will be non-positive. As a consequence, the
Hicksian demand function exists and satisfies Shephard’s identity almost everywhere in p, without
any requirement that the utility function be quasi-concave.

Consider budgets(pN,yN) and (pO,yO), and the convex combination p* = 2pN+(1-2)pO and y*
= 2yN+(1-2)yO with 0 < 2 < 1. We now demonstrate that V(p*,y*) # max{V(pN,yN),V(pO,yO)}. A
function with this property is said to be quasi-convex. Suppose x* maximizes utility subject to the
constraint p*@x # y*, so that V(p*,y*) = U(x*). Then 2(pN@x*-yN) +(1-2)pO@x*-yO) # 0, implying
that either pN@x* # yN and hence U(x*) # V(pN,yN), or else pO@x* # yO and hence U(x*) # V(pO,yO).
Therefore, V(p*,y*) = U(x*) # max{V(pN,yN),V(pO,yO)}, establishing the result.

From the identity y / M(p,V(p,y)) and the result H(p,u) = LpM(p,u), one can take derivatives
with respect to y and p to obtain

1 / LuM(p,u)) . u=V(p,y) Vy(p,y) and 0 / LuM(p,u) . u=V(p,y)@LpV(p,y) + LpM(p,u) . u=V(p,y).

6
Then, using the first of these equations to eliminate the term LuM(p,u) . u=V(p,y) in the second gives
a relationship between the indirect utility function and the Marshallian demand functions that is
termed Roy’s identity,

X1(p,y) / H1(p,V(p,y)) / MM(p,u)/Mp1 . u=V(p,y) / -(MV(p,y)/Mp1)/(MV(p,y)/My).

From the identity X1(p,y) / H1(p,V(p,y)), by taking derivatives one obtains the relationships
LyX1(p,y) / LuH1(p,u). u=V(p,y)LyV(p,y) and LpX1(p,y) / LuH1(p,u). u=V(p,y)LpV(p,y) + LpH1(p,u). u=V(p,y).
Use the first equation to obtain LuH1 = LyX1/LyV, and then substitute this into the second equation
to obtain

LpX1(p,y) / [LpV(p,y)/LyV(p,y)]LyX1(p,y) + LpH1(p,u) . u=V(p,y)

/ - X(p,y)LIX1(p,y) + LpH1(p,u) . u=V(p,y).

Written out, this equation decomposes the overall effect of own price on market demand into a
substitution effect MH1(p,u)/Mp1 . u=V(p,y) = M2M(p,u)/Mp12 . u=V(p,y), which corresponds to the change in
demand when income is compensated to keep the consumer on the same indifference curve, and
is always non-positive, and an income effect, which for the own price change is - X1(p,y)LyX1(p,y).
If commodity 1 is normal, its change with income, LyX1(p,y), is non-negative, and the income effect
reinforces the non-positive substitution effect so that demand for commodity 1 falls when its price
rises, and the “law of demand” holds. If commodity 1 is inferior, its change with income is
negative, and the income effect tends to offset the non-positive substitution effect. Usually, the
income effect for an inferior good is not strong enough to completely offset the substitution effect.

An interesting variant of the classical consumer demand problem occurs when instead of
considering fixed income y when prices change, one considers income y = A(p,t) that is a function
of prices and of policies (denoted by t) that control taxes and transfers which can alter income. For
example, one might consider a consumer who has an initial endowment of goods T, and receives
income A(p,t) = p@T from selling this endowment in the market. Endowments of leisure that are
sold as labor are a leading case. If one decomposes the effect on demand of a change in price into
substitution and income effects, one gets as before a substitution effect that never increases demand
for a good when its price rises, but now the income effect has the pattern given in the following
table:

Income Effect Consumer Net Demander of Good Consumer Net Supplier of Good
Normal Good Reinforces substitution effect Offsets substitution effect
Inferior Good Offsets substitution effect Reinforces substitution effect