1 Example: Cobb-Douglas utility

Cobb-Douglas utility1 is dened as

u= xi i , or, after monotonous transformation, u = i ln xi ,
i i

with e = 1 (i.e. the i s sum up to one).

1.1 Marshallian Demand

Optimal consumption (Marshallian demand) solves

max u(x) s.t. px = y.

It can be derived by the Lagrange-method. The Lagrangian L,

L = u(x) + (y px),

is dierentiated with respect to xi and , which yields the FOC:2

dL u(x)
i : = pi = 0,
dxi xi

and the budget line, px = y. Using the log-representation, we get

dL i i
i : = = pi = pi xi .
dxi xi

Summation over i gives

1 1
i = pi xi , or = y.

i i

Substituting y 1 for in the FOCs gives

i : i y = pi xi , (1)

1
and nally the Marshallian demand functions

y
i : xi = i . (2)
pi

Note that (1) gives a key-implication of Cobb-Douglas utility on optimal consumption: The

income shares spent on the various commodities are constant and given by i .

1.2 Indirect utility and Roys identity

The indirect utitility function results from plugging (2) into the utility function,

( )
( 1 )i
v(p, y) xi (p, y) i
=y ii . (3)
pi
i i i

Roys identity allows to recover Marshallian demand from v(p, y) by

v(p,y)
pi
x(p, y) = v(p,y) . (4)
y

Applying this to (3) gives

( )i 1 ( )2 ( ) ( )
v(p, y) 1 1 1 j
= i ii y, (5)
pi pi pi pj
i j=i

or

( ) ( )(
( 1 )i
)
v(p, y) 1
= i ii y. (6)
pi pi pi
i i

Since

( )
v(p, y) ( 1 )i
= ii , (7)
y pi
i i

we get Marshallian demand again.

2
1.3 Expenditure function and Shephards Lemma

The expenditure function is dened as

e(p, u) min px s.t. u(x) = u,

where prices p and utility u are given. First note that e is concave in prices, i.e.

e(tp1 + (1 t)p2 , u) te(p1 , u) + (1 t)e(p2 , u), (8)

for all p1 , p2 and t [0, 1] and given u > 0. In particular, 2 e(p, u)/p2i 0. The proof is

straightforward:

Dene p = tp1 + (1 t)p2 for an arbitrary t [0, 1]. Let x1 , x2 , x denote the expenditure min-

imizing consumptions for given prices p1 , p2 , p . By denition, p1 x1 p1 x and p2 x2 p2 x .

Multiplying both sides of p1 x1 p1 x with t and p2 x2 p2 x with 1 t gives

tp1 x1 tp1 x and (1 t)p2 x2 (1 t)p2 x .

Adding up both left hand sides and right hand sides leads to

tp1 x1 + (1 t)p2 x2 tp1 x + (1 t)p2 x .

But this is just

te(p1 , u) + (1 t)e(p2 , u) (tp1 + (1 t)p2 )x = e(p , u).

For the Cobb-Douglas utility function, we use (3) to give the expenditure function. Note that

u = v(p, y), hence solving for y gives the expenditure function:

( )
( 1 )i
e(p, u) = y = u (pi )i . (9)
i
i i

3
Applying Shephards Lemma,

e(p, u)
= xh (p, u), (10)
pi

to (9) gives

( )
i ( 1 )i
h
x (p, u) = u (pi )i . (11)
pi i
i i

Notes
1
Named after Charles W. Cobb and Paul H. Douglas, who published an econometric analysis of the relation
between labour, capital and output in AER 1928. They used this type of specication.
2
FOC: rst order optimality conditions.