PROFIT FUNCTIONS

1. REPRESENTATION OF TECHNOLOGY
1.1. Technology Sets. The technology set for a given production process is dened as
T = {(x, y) : x R
n
+
, y R
m:
+
x can produce y}
where x is a vector of inputs and y is a vector of outputs. The set consists of those combinations
of x and y such that y can be produced from the given x.
1.2. The Output Correspondence and the Output Set.
1.2.1. Denitions. It is often convenient to dene a production correspondence and the associated
output set.
1: The output correspondence P, maps inputs x R
n
+
into subsets of outputs, i.e., P: R
n
+

2
R
m
+
. A correspondence is different from a function in that a given domain is mapped into
a set as compared to a single real variable (or number) as in a function.
2: The output set for a given technology, P(x), is the set of all output vectors y R
m
+
that are
obtainable from the input vector x R
n
+
. P(x) is then the set of all output vectors y R
m
+
that are obtainable from the input vector x R
n
+
. We often write P(x) for both the set based
on a particular value of x, and the rule (correspondence) that assigns a set to each vector x.
1.2.2. Relationship between P(x) and T(x,y).
P (x) = (y : (x, y ) T)
1.2.3. Properties of P(x).
P.1a: P.1 No Free Lunch. 0 P(x) x R
n
+
.
P.1b: y P(0), y > 0.
P.2: Input Disposability. x R
n
+
, P(x) P(x), 1.
P.2.S: Strong Input Disposability. x, x R
n
+
, x x P(x) P(x).
P.3: Output Disposability. x R
n
+
, y P(x) and 0 1 y P(x).
P.3.S: Strong Output Disposability. x R
n
+
, y P(x) y P(x), 0 y y.
P.4: Boundedness. P(x) is bounded for all x R
n
+
.
P.5: T(x) is a closed set P: R
n
+
2
R
m
+
is a closed correspondence, i.e., if [x

x
0
, y

y
0
and
y

P(x

), ] then y
0
P(x
0
).
P.6: Attainability. If y P(x), y 0 and x 0, then 0,

0 such that y P(

x).
P.7: P(x) is convex
P(x) is convex for all x R
n
+
.
P.8: P is quasi-concave.
The correspondence P is quasi-concave on R
n
+
which means x, x R
n
+
, 0 1, P(x)
P(x) P(x + (1-)x)
P.9: Convexity of T(x). P is concave on R
n
+
which means x, x R
n
+
, 0 1, P(x)+(1-
)P(x) P(x + (1-)x)
Date: September 14, 2005.
1
2 PROFIT FUNCTIONS
1.2.4. Properties of T(x,y).
T.1a: Inaction and No Free Lunch. (0,y) T x R
n
+
and y R
m
+
. This implies that T(x,y) is a
non-empty subset of R
m+n
+
T.1b: (0,y) T, y 0, y = 0.
T.2: Input Disposability. If (x,y) T and 1 then (x, y) T.
T.2.S: Strong Input Disposability. If (x,y) T and x x, then (x, y) T.
T.3: Output Disposability. (x,y) R
n+m
+
,if (x, y) T and 0 < 1 then (x, y) T.
T.3.S: Strong Output Disposability. If (x,y) T and y y, then (x, y) T.
T.4: Boundedness. For every nite input vector x 0, the set y P(x) is bounded from above.
This implies that only nite amounts of output can be produced from nite amounts of
inputs.
T.5: T(x) is a closed set. The assumption that P(x) and V(y) are closed does not imply that T
is a closed set, so it is assumed. Specically, if [x

x
0
, y

y
0
) and (x

, y

) T, ] then
(x
0
y
0
) T.
T.9: T is a convex set. This is not implied by the convexity of P(x) and V(y). Specically, a
technology could exhibit a diminishing rate of technical substitution, a diminishing rate of
product transformation and still exhibit increasing returns to scale.
2. PROFIT MAXIMIZATION AND THE PROFIT FUNCTION
2.1. Prot Maximization.
2.1.1. Setup of problem. The general rm-level maximization problem can be written in a number of
alternative ways.
= max
x y
[
m

j=1
p
j
y
j

n

i=1
w
i
x
i
], such that (x, y) T. (1)
where T is represents the graph of the technology or the technology set. The problem can also
be written as
= max
x y
[
m

j=1
p
j
y
j

n

i=1
w
i
x
i
] such hat x V(y) ] (2a)
= max
x y
[
m

j=1
p
j
y
j

n

i=1
w
i
x
i
] such that y P(x) ] (2b)
where the technology is represented by V(y), the input requirement set, or P(x),the output set.
Though T is non-empty, closed and convex, prot may not attain a maximumon T, i.e. prot can be
unbounded even when the technology set is well behaved if output price is higher than input price.
Consider for example the production function f(x)= x + x
1/2
. The production function is concave
but prot will go to innity if p > w. Therefore we often write
= sup
x y
[
m

j=1
p
j
y
j

n

i=1
w
i
x
i
], such that (x, y) T. (3)
where sup stands for supremum and could be innity. We understand that in all practical prob-
lems the sup is a max.
PROFIT FUNCTIONS 3
If we carry out the maximization in equation 1 or equation 3, we obtain a vector of optimal
outputs and a vector of optimal outputs such that y is producible given x and prots cannot be
increased. We denote these optimal input and output choices as y(p,w) and x(p,w) where it is
implicit that y and x are vectors.
2.1.2. One output and two input example. In the case of a single output (such that f(x) is dened) and
two inputs we obtain the following.
= p f (x
1
, x
2
) w
1
x
1
w
2
x
2
(4)
If we differentiate the expression in equation 4 with respect to each input we obtain

x
1
= p
f (x
1
, x
2
)
x
1
w
1
= 0

x
2
= p
f (x
1
, x
2
)
x
2
w
2
= 0
(5)
If we solve the equations in 5 for x
1
and x
2
, we obtain the optimal values of x for a given p and
w. As a function of w for a xed p, this gives the vector of factor demands for x.
x

= x(p, w
1
, w
2
)
= ( x
1
(p, w
1
, w
2
), x
2
(p, w
1
, w
2
) )
(6)
The optimal output is given by
y

= f ( x
1
(p, w
1
, w
2
), x
2
(p, w
1
, w
2
) ) (7)
2.2. The Prot Function. If we substitute the optimal input demand from equation 6 into equation
1 or equation 4, we obtain the prot function. The prot function is usually designated by .
(p, w) =
m

j=1
p
j
y
j
(p, w)
n

i=1
w
i
x
i
(p, w)
= p f(x
1
(p, w), x
2
(p, w)) w
1
x
1
(p, w) w
2
x
2
(p, w)
(8)
Notice that is a function of p and w, not x or y. The optimal x and optimal y have already been
chosen. The function tells us what prots will be (assuming the rm is maximizing prots) given a
set of output and input prices.
To help understand how (p,w) only depends on p and w, consider the prot function for the
case of two inputs.
(p, w
1
, w
2
) = p f(x
1
(p, w
1
, w
2
), x
2
(p, w
1
, w
2
)) w
1
x
1
(p, w
1
, w
2
) w
2
x
2
(p, w
2
, w
2
) (9)
Consider the derivative of (p,w) with respect to p.
(p, w)
p
= p
f(x
1
(p, w), x
2
(p, w))
x
1
x
1
(p, w)
p
+ p
f(x
1
(p, w), x
2
(p, w))
x
2
x
2
(p, w)
p
+ f(x
1
(p, w), x
2
(p, w)) w
1
x
1
(p, w))
p
w
2
x
2
(p, w))
p
(10)
Now collect terms containing
x1(p,w)
p
and
x2(p,w)
p
.
4 PROFIT FUNCTIONS
(p, w)
p
=
x
1
(p, w)
p
_
p
f(x
1
(p, w), x
2
(p, w))
x
1
w
1
_
+
x
2
(p, w))
p
_
p
f(x
1
(p, w), x
2
(p, w))
x
2
w
2
_
+ f(x
1
(p, w), x
2
(p, w))
(11)
But the rst order conditions in equation 5 imply that
p
f(x
1
x
2
)
x
i
w
i
= 0, i = 1, 2 (12)
This means that the two bracketed terms in equation 11 are equal to zero so that
(p, w)
p
= f(x
1
(p, w), x
2
(p, w)) (13)
which depends only on p and w.
3. PROPERTIES OF, OR CONDITIONS ON, THE PROFIT FUNCTION
3.1. .1. (p, w) is an extended real valued function (it can take on the value of +for nite prices)
dened for all (p, w) (0
m
, 0
n
)) and (p, w) pa - wb for a xed vector (a, b) (0
m
, 0
n
). This
implies that (p, w) 0 if (0
m
, 0
n
) T(x,y),which we normally assume.
3.2. .2. is nonincreasing in w
3.3. .3. is nondecreasing in p
3.4. .4. is a convex function
3.5. .5. is homogeneous of degree 1 in p and w.
3.6. .6 Hotellings Lemma.
(p, w)
p
j
= y
j
(p, w) (14a)
(p, w)
w
i
= x
i
(p, w) (14b)
4. DISCUSSION OF PROPERTIES OF THE PROFIT FUNCTION
For ease of exposition, consider the case where there is a single output. This is easily generalized
by replacing f(x(p,w)) with y(p,w) and letting p be a vector.
4.1. .1. (p, w) is an extended real valued function (it can take on the value of +for nite prices)
dened for all (p, w) (0
m
, 0
n
)) and (p, w) pa - wb for a xed vector (a, b) (0
m
, 0
n
). This
implies that (p, w) 0.
The prot function can be innity due to the fact that maximum prots may be innity as dis-
cussed in section 2.1.1. (p, w) pa - wb for a xed vector (a, b) because (p, w) is the maximum
prot at output prices p and input prices w. Any other input combination is bounded by (p, w).
PROFIT FUNCTIONS 5
4.2. .2. is nonincreasing in w
Let optimal input be x(p,w) with prices p and w and x(p,w) with prices p and w. Now assume
that w > w. It is clear that pf( x(p,w) ) - w x(p,w) pf( x(p,w)) - w x(p,w) because the optimal x with
w is x. However, pf( x) - w x < pf( x ) - w x because w > w by assumption. So we obtain
pf( x(p, w) ) w x(p, w) < pf( x(p, w)) w x(p, w) pf( x(p, w) ) w(p, w) x(p, w)
pf( x(p, w) ) w x(p, w) pf( x(p, w) ) w x(p, w)
(15)
4.3. .3. is nondecreasing in p
Let prot be ( p,w) with prices p and w and ( p,w) with prices p and w. Now assume that p >
p. It is clear that
pf( x(p, w) ) w x(p, w) p f( x(p, w)) w x(p, w) (16)
because x(p, w) is optimal for prices p. However,
pf( x(p, w) ) w x(p, w) < p f( x(p, w)) w x(p, w) (17)
because p > p by assumption. So we obtain,
pf( x(p, w) ) w x(p, w) < p f( x(p, w)) w x(p, w) p f( x(p, w)) w x(p, w)
pf( x(p, w) ) w x(p, w) p f( x(p, w)) w x(p, w)
(18)
4.4. .4. is a convex function
Consider the denition of a convex function. A function f: R
m+n
R
1
is said to be convex if
f ( x
1
+ (1 ) x
2
) f (x
1
) + (1 )f ( x
2
) for each vector x
1
, x
2
R
m+n
and for each
[0, 1]. In the case of the prot function maps an n+m vector containing output and input prices
into the real line. Alternatively, a differentiable function is convex in R
m+n
+
if and only if
(p, w) ( p, w) +
p
( p, w)

(p p) +
w
( p, w)

(w w),
for each distinct (p, w), ( p, w) R
m+n
+
.
(19)
Let (y,x) be the prot maximizing choices of y and x when prices are(p, w) and let (y,x) be
prot maximizing maximizing choices of y and x when prices are(p, w). Let (p w) be a linear
combination of the price vectors (p,w) and (p,w), i.e.
(p

, w

) = (p, w) + (1 ) (p

, w

)
= [ p + (1 ) p

, w + (1 ) w

]
(20)
Then let (y,x) be the prot maximizing choices of y and x when prices are(p, w). This then
implies that
(p

, w

) = p

(21)
Now substitute the denition of p and w from equation 20 into 21
6 PROFIT FUNCTIONS
(p

, w

) = p

= [ p + (1 ) p

] y

[ w + (1 ) w

] x

= p y

wx

+ (1 ) p

(1 ) w

= [ p y

wx

] + (1 ) [ p

]
(22)
Be the denition of prot maximization, we know that when prices are (p,w) any output input
combination other than (y,x) will yield lower prots. Similarly when prices are (p, w). We then
can write
[ p y

wx

] [ p y wx] = (p, w)
(1 ) [ p

] (1 ) [ p

] = (1 ) (p

, w

)
(23)
Now write equation 23 without the middle terms
[ p y

wx

] (p, w)
(1 ) [ p

] (1 ) (p

, w

)
(24)
Now add the two inequalities in equation 24 to obtain
[ p y

wx

] + (1 ) [ p

] (p, w) + (1 ) (p

, w

) (25)
Now substitute for the left hand side of equation 25 from equation 22 to obtain
(p

, w

) (p, w) + (1 ) (p

, w

) (26)
Becauuse (p,w) is convex, we knowthat its Hessian matrix is positive semidenite. This means
that the diagonals of the Hessian matrix are all positive or zero, i.e.,

2
(p,w)
p
2
j
0 and

2
(p,w)
w
2
i
0.
We can visualize convexity if we hold all input prices xed and only consider a rm with a single
output, or hold all but one output price xed. In gure 1we can see that the tangent lies below the
curve. At prices (p

, w

), prots are at the level (p

, w

). If we hold the output and input levels

xed at (y

, x

) and change p, we move along the tangent line denoted by

(p

, w). Prots are less

along this line than along the prot function because we are not adjusting x and y to account for
the change in p.
4.5. .5. is homogeneous of degree 1 in p and w.
Consider the prot maximization problem,
(p, w) = max
x y
[
m

j=1
p
j
y
j

n

i=1
w
i
x
i
], such that (x, y) T. (27)
Now multiply all prices by and denote the new prot mximization problem as
PROFIT FUNCTIONS 7
FIGURE 1. The Prot Function is Convex
p
Profits
p,w

,w

py

,w

(p, w) = max
x y
[
m

j=1
p
j
y
j

n

i=1
w
i
x
i
], such that (x, y) T
= max
x y
[
m

j=1
p
j
y
j

n

i=1
w
i
x
i
]

(p, w) = (p, w)
(28)
4.6. .6 Hotellings Lemma.
(p, w)
p
j
= y
j
(p, w) (29a)
(p, w)
w
i
= x
i
(p, w) (29b)
We have already shown part a of equation 29 in equation 13 for the case of a single output. We
can derive part b of f equation 29 in a similar manner. Consider the derivative of (p,w
1
, w
2
) with
respect to w
1
. The prot function is given by
(p, w
1
, w
2
) = p f(x
1
(p, w
1
, w
2
), x
2
(p, w
1
, w
2
)) w
1
x
1
(p, w
1
, w
2
) w
2
x
2
(p, w
2
, w
2
) (30)
Taking the derivative with respect to w
1
we obtain
(p, w)
w
1
= p
f(x
1
(p, w), x
2
(p, w))
x
1
x
1
(p, w)
w
1
+ p
f(x
1
(p, w), x
2
(p, w))
x
2
x
2
(p, w)
w
1
w
1
x
1
(p, w))
w
1
x
1
(p, w) w
2
x
2
(p, w))
w1
(31)
Now collect terms containing
x1(p,w)
w1
and
x2(p,w)
w1
.
8 PROFIT FUNCTIONS
(p, w)
w
1
=
x
1
(p, w)
w
1
_
p
f(x
1
(p, w), x
2
(p, w))
x
1
w
1
_
+
x
2
(p, w))
w
1
_
p
f(x
1
(p, w), x
2
(p, w))
x
2
w
2
_
x
1
(p, w
1
, w
2
)
(32)
But the rst order conditions in equation 5 imply that
p
f(x
1
x
2
)
x
i
w
i
= 0, i = 1, 2 (33)
This means that the two braketed terms in equation 32 are equal to zero so that
(p, w)
w1
= x
1
(p, w
1
, w
2
) (34)
We can show this in a more general manner as follows. Dene a function g(p,w) as follows
g(p, w) = (p, w) [ py

wx

] (35)
where (y

, x

) is some production plan that is not necessarily optimal at prices (p,w). At prices
(p

, w

), (y

, x

) will be the prot maximizing production. The rst term in equation 35 will always
be greater than or equal to the second term. Thus g(p,w) will reach its minimum when prices are
(p

, w

). Consider then the conditions for minimizing g(p,w).

g(p

, w

)
p
j
=
(p

, w

)
p
j
y

j
= 0, j = 1, 2, . . ., m
g(p

, w

)
w
i
=
(p

, w

)
w
i
+ x

j
= 0, i = 1, 2, . . . , n
(36)
Rearranging equation 36 we obtain
(p

, w

)
p
j
y

j
= 0, j = 1, 2, . . ., m

(p

, w

)
p
j
= y

j
, j = 1, 2, . . . , m
(p

, w

)
w
i
+ x

j
= 0, i = 1, 2, . . . , n

(p

, w

)
w
i
= x

i
, i = 1, 2, . . ., m
(37)
A number of important implications come from equation 37.
1: We can obtain output supply and input demand equations by differentiating the prot
function.
2: If we have an expression (equation) for the prot function, we can obtain output supply
and input demand equations or functional forms for such equations by differentiating the
prot function as compared to solving a maximization problem. This would allow one to
nd functional forms for estimating supply and demand without solving maximization
problems.
PROFIT FUNCTIONS 9
3: Given that the output supply and input demand functions are derivatives of the prot
function, we can determine many of the properties of these response functions by under-
standing the properties of the prot function. For example, because the prot function is
homogeneous of degree one in p and w, its derivatives are homogeneous of degree zero.
This means that output supply and input demand equations are homogeneous of degree
zero, i.e., multiplying all prices by the same constant will not change output supply or
input demand.
4: The second derivatives of the prot function are the rst derivatives of the output supply
and input demand functions. Properties of the second derivatives of the prot function
are properties of the rst derivatives of the output supply and input demand functions.
5: The symmetry of second order cross partial derivatives leads to symmetry of rst cross
price derivatives of the output supply and input demand functions.
5. NUMERIC EXAMPLE
5.1. Production function. Consider the following production function.
y = f(x
1
, x
2
)
= 24x
1
+ 14x
2
x
2
1
+ x
1
x
2
x
2
2
(38)
The rst and second partial derivatives are given by
f(x
1
, x
2
)
x
1
= 24 2x
1
+ x
2
f(x
1
, x
2
)
x
2
= 14 + x
1
2x
2

2
f(x
1
, x
2
)
x
2
1
= 2

2
f(x
1
, x
2
)
x
1
x
2
= 1

2
f(x
1
, x
2
)
x
2
2
= 2
(39)
The Hessian is

2
f(x
1
, x
2
) =
_
2 1
1 2
_
(40)
The determinant of the Hessian is given by

2 1
1 2

= 4 1 = 3 (41)
The rate of technical substitution is given by
RTS =
x
2
(y, x
1
)
x
1
=
24 + 2x
1
x
2
14 + x
1
2x
2
(42)
The elasticity substitution is given by
10 PROFIT FUNCTIONS

12
=
f
1
f
2
(x
1
f
1
+ x
2
f
2
)
x
1
x
2
(f
11
f
2
2
2f
12
f
1
f
2
+ f
22
f
2
1
)
=
(14 + x
1
2x
2
)(24 2x
1
+ x
2
) (x
2
1
12x
1
7x
2
x
1
x
2
+ x
2
2
)
x
1
x
2

1108 + 3x
2
1
72x
1
42x
2
3x
1
x
2
+ 3x
2
2
)

(43)
5.2. Prot maximization. Prot is given by
= p f(x
1
, x
2
) w
1
x
1
w
2
x
2
= p
_
24x
1
+ 14x
2
2x
2
1
+ x
1
x
2
2x
2
2

w
1
x
1
w
2
x
2
(44)
We maximize prot by taking the derivatives of 44 setting them equal to zero and solving for x
1
and x
2
.

x
1
= p [ 24 2x
1
+ x
2
] w
1
= 0

x
2
= p [ 14 + x
1
2x
2
] w
2
= 0
(45)
Rearranging 45 we obtain
24 2x
1
+ x
2
=
w
1
p
(46a)
14 + x
1
2x
2
=
w
2
p
(46b)
Now solve equation 46a for x
1
as follows
x
1
=
24p w
1
+ px
2
2p
(47)
Then substitute x
1
from equation 47 into equation 46b as follows
PROFIT FUNCTIONS 11
14 +
_
24p w
1
+ px
2
2p
_
2x
2
=
w
2
p

_
28p + 24p w
1
+ px
2
4px
2
2p
_
=
w
2
p

_
52p w
1
3px
2
2p
_
=
w
2
p

_
3px
2
2p
_
=
w
2
p

_
52p w
1
2p
_
=
_
2w
2
52p + w
1
2p
_
3px
2
= 2w
2
52p + w
1
x
2
=
2w
2
+ 52p w
1
3p
=
52p w
1
2w
2
3p
(48)
If we substitute the last expression in equation 48 for x
2
in equation 47 we obtain
x
1
=
24p w
1
+ px
2
2p
=
24p w
1
2p
+
p
2p
x
2
=
24p w
1
2p
+
1
2
_
52p w
1
2w
2
3p
_
=
72p 3w
1
+ 52p w
1
2w
2
6p
=
124p 4w
1
2w
2
6p
=
62p 2w
1
w
2
3p
(49)
5.3. Necessary and sufcient conditions for a maximum. Consider the Hessianof the prot equa-
tion.

2
(x
1
, x
2
) = p
_
2 1
1 2
_
=
_
2p p
p 2p
_
(50)
For a maximumwe need the diagonal elements to be negative and the determinant to be positive.
The diagonal elements are negative. The determinant of the Hessian is 4p
2
-p
2
= 3p
2
, which is
positive.
12 PROFIT FUNCTIONS
5.4. Optimal output. We can obtain optimal output as a function of p, w
1
and w
2
by substituting
the optimal values of x
1
and x
2
into the production function.
y(p, w
1
, w
2
) = f(x
1
(p, w
2
, w
2
), x
2
(p, w
1
, w
2
))
= 24
_
62p 2w
1
w
2
3p
_
+ 14
_
52p w
1
2w
2
3p
_

_
62p 2w
1
w
2
3p
_
2
+
_
62p 2w
1
w
2
3p
__
52p w
1
2w
2
3p
_

_
52p w
1
2w
2
3p
_
2
(51)
Now put everything over a common denominator and multiply out as follows
y(p, w
1
, w
2
) =
(72p) (62p 2w
1
w
2
)
9p
2
+
42p (52p w
1
2w
2
)
9p
2

(62p 2w
1
w
2
)
2
9p
2
+
(62p 2w
1
w
2
) (52p w
1
2w
2
)
9p
2

(52p w
1
2w
2
)
2
9p
2
=
4464p
2
144pw
1
72pw
2
9p
2
(52a)
+
2184p
2
42pw
1
84pw
2
9p
2
(52b)

3844p
2
248pw
1
+ 4w
2
1
124pw
2
+ 4w
1
w
2
+ w
2
2
9p
2
(52c)
+
3224p
2
166pw
1
+ 2w
2
1
176pw
2
+ 5w
1
w
2
+ 2w
2
2
9p
2
(52d)

2704p
2
104pw1 + w
2
1
208pw
2
+ 4w
1
w
2
+ 4w
2
2
9p
2
(52e)
Notice that we have terms in the following p
2
, pw
1
pw
2
, w
2
1
, w
1
w
2
and w
2
2
. Now rearrange
terms in equation 52 combining like terms.
PROFIT FUNCTIONS 13
y(p, w
1
, w
2
) =
4464p
2
+ 2184p
2
3844p
2
+ 3224p
2
2704p
2
9p
2
(53a)
+
144pw
1
42pw
1
+ 248pw
1
166pw
1
+ 104pw
1
9p
2
(53b)
+
72pw
2
84pw
2
+ 124pw
2
176pw
2
+ 208pw
2
9p
2
(53c)
+
4w
2
1
+ 2w
2
1
w
2
1
9p
2
(53d)
+
4w
1
w
2
+ 5w
1
w
2
4w
1
w
2
9p
2
(53e)
+
w
2
2
+ 2w
2
2
4w
2
2
9p
2
(53f)
Now combine terms
y(p, w
1
, w
2
) =
3324p
2
9p
2
+
0pw
1
9p
2
+
0pw
2
9p
2
+
3w
2
1
9p
2
+
3w
1
w
2
9p
2
+
3w
2
2
9p
2
=
1108p
2
3p
2

w
2
1
3p
2

w
1
w
2
3p
2

w
2
2
3p
2
=
1108
3

w
2
1
3p
2

w
1
w
2
3p
2

w
2
2
3p
2
(54)
5.5. The example numerical prot function. We obtain the prot function by substituting the op-
timal values of x
1
and x
2
into the prot equation
(p, w
1
, w
2
) = pf(x
1
(p, w
2
, w
2
), x
2
(p, w
1
, w
2
)) w
1
x
1
(p, w
2
, w
2
) w
2
x
2
(p, w
2
, w
2
)
= p
_
1108p
2
3p
2

w
2
1
3p
2

w
1
w
2
3p
2

w
2
2
3p
2
_
w
1
_
62p 2w
1
w
2
3p
_
w
2
_
52p w
1
2w
2
3p
_
=
1108p
2
w
2
1
w
1
w
2
w
2
2
62pw
1
+ 2w
2
1
+ w
1
w
2
52pw
2
+ w
1
w
2
+ 2w
2
2
3p
=
1108p
2
+ w
2
1
+ w
2
2
+ w
1
w
2
62pw
1
52pw
2
3p
(55)
5.6. Optimal input demands via Hotellings lemma. We can nd the optimal input demands by
taking the derivative of equation 86 with respect to w
1
and w
2
. First with respect to w
1
14 PROFIT FUNCTIONS
(p, w
1
, w
2
) =
1108p
2
+ w
2
1
+ w
2
2
+ w
1
w
2
62pw
1
52pw
2
3p
(p, w
1
, w
2
)
w
1
=
2w
1
+ w
2
62p
3p
x
1
(p, w
1
, w
2
) =
2w
1
+ w
2
62p
3p
=
62p 2w
1
w
2
3p
(56)
Then with respect to w
2
(p, w
1
, w
2
)
w
2
=
2w
2
+ w
1
52p
3p
x
2
(p, w
1
, w
2
) =
2w
2
+ w
1
52p
3p
=
52p w
1
2w
2
3p
(57)
5.7. Optimal output via Hotellings lemma. Take the derivative of the prot function with respect
to p using the quotient rule
(p, w
1
, w
2
) =
1108p
2
+ w
2
1
+ w
2
2
+ w
1
w
2
62pw
1
52pw
2
3p
(p, w
1
, w
2
)
p
=
3p (2216p 62w
1
52w
2
) 3
_
1108p
2
+ w
2
1
+ w
2
2
+ w
1
w
2
62pw
1
52pw
2
_
9p
2
=
6648p
2
186pw
1
156pw
2
3324p
2
3w
2
1
3w
2
2
3w
1
w
2
+ 186pw
1
+ 156pw
2
9p
2
=
3324p
2
3w
2
1
3w
2
2
3w
1
w
2
9p
2
y(p, w
1
, w
2
) =
1108p
2
w
2
1
w
2
2
w
1
w
2
3p
2
(58)
6. ALGEBRAIC EXAMPLE
6.1. Production function. Consider the following production function.
y = f(x
1
, x
2
)
=
1
x
1
+
2
x
2
+
11
x
2
1
+
12
x
1
x
2
+
22
x
2
2
(59)
The rst and second partial derivatives are given by
PROFIT FUNCTIONS 15
f(x
1
, x
2
)
x
1
=
1
+ 2
11
x
1
+
12
x
2
f(x
1
, x
2
)
x
2
=
2
+
12
x
1
+ 2
22
x
2

2
f(x
1
, x
2
)
x
2
1
= 2
11

2
f(x
1
, x
2
)
x
1
x
2
=
12

2
f(x
1
, x
2
)
x
2
2
= 2
22
(60)
The Hessian is

2
f(x
1
, x
2
) =
_
2
11

12

12
2
22
_
(61)
The determinant of the Hessian is given by

2
11

12

12
2
11

= 4
11

22

2
12
(62)
The rate of technical substitution is given by
RTS =
x
2
(y, x
1
)
x
1
=

1
+ 2
11
x
1
+
12
x
2

2
+
12
x
1
+ 2
22
x
2
(63)
The elasticity substitution is given by

12
=
f
1
f
2
(x
1
f
1
+ x
2
f
2
)
x
1
x
2
(f
11
f
2
2
2f
12
f
1
f
2
+ f
22
f
2
1
)
=
(
1
+ 2x
1

11
+ x
2

12
)(
2
+ x
1

12
+ 2x
2

22
) (2x
2
1

11
+ x
1
(
1
+ 2x
2

12
) + x
2
(
2
+ 2x
2

22
))
x
1
x
2
(2 (
1
+ 2x
1

11
+ x
2

12
)
2

22
2
12
(
1
+ 2x
1

11
+ x
2

12
) (
2
+ x
1

12
+ 2x
2

22
) + 2
11
(
2
+ x
1

12
+ 2x
2

22
)
2
))
(64)
6.2. Prot maximization. Prot is given by
= p f(x
1
, x
2
) w
1
x
1
w
2
x
2
= p
_

1
x
1
+
2
x
2
+
11
x
2
1
+
12
x
1
x
2
+
22
x
2
2

w
1
x
1
w
2
x
2
(65)
We maximize prot by taking the derivatives of 65 setting them equal to zero and solving for x
1
and x
2
.

x
1
= p [
1
+ 2
11
x
1
+
12
x
2
] w
1
= 0

x
2
= p [
2
+
12
x
1
+ 2
22
x
2
] w
2
= 0
(66)
Rearranging 66 we obtain
16 PROFIT FUNCTIONS

1
+ 2
11
x
1
+
12
x
2
=
w
1
p
(67a)

2
+
12
x
1
+ 2
22
x
2
=
w
2
p
(67b)
Now solve equation 67a for x
1
as follows
x
1
=
w
1
p
1
px
2

12
2p
11
(68)
Then substitute x
1
from equation 68 into equation 67b as follows

2
+
12
_
w
1
p
1
px
2

12
2p
11
_
+ 2
22
x
2
=
w
2
p

2p
2

11
+ w
1

12
p
1

12
px
2

2
12
+ 4px
2

11

22
2p
11
=
w
2
p

2p
2

11
+ w
1

12
p
1

12
2p
11
+
4px
2

11

22
px
2

2
12
2p
11
=
w
2
p

2p
2

11
+ w
1

12
p
1

12
2p
11
+
x
2
_
4
11

22

2
12
_
2
11
=
w
2
p
(69)
Now isolate x
2
on the left hand side of equation 69 as follows.
x
2
_
4
11

22

2
12
_
2
11
=
w
2
p

2p
2

11
+ w
1

12
p
1

12
2p
11
x
2
=
2
11
w
2
p (4
11

22

2
12
)

(2
11
) (2p
2

11
+ w
1

12
p
1

12
)
2p
11
(4
11

22

2
12
)
=
2
11
w
2
p (4
11

22

2
12
)

(2p
2

11
+ w
1

12
p
1

12
)
p (4
11

22

2
12
)
=
2
11
w
2
2p
2

11
w
1

12
+ p
1

12
p (4
11

22

2
12
)
=
2
11
w
2
w
1

12
+ p (
1

12
2
2

11
)
p (4
11

22

2
12
)
(70)
If we substitute the last expression in equation 70 for x
2
in equation 68 we obtain
x
1
=
w
1
p
1
px
2

12
2p
11
=
w
1
p
1
2p
11

p
12
2p
11
x
2
=
w
1
p
1
2p
11

p
12
2p
11
_
2
11
w
2
2p
2

11
w
1

12
+ p
1

12
p (4
11

22

2
12
)
_
(71)
Now put both terms in equation 71 over a common denominator and simplify
PROFIT FUNCTIONS 17
x
1
=
p
_
4
11

22

2
12
_
(w
1
p
1
) p
12
(2
11
w
2
2p
2

11
w
1

12
+ p
1

12
)
2p
11
p (4
11

22

2
12
)
=
_
4
11

22

2
12
_
(w
1
p
1
)
12
(2
11
w
2
2p
2

11
w
1

12
+ p
1

12
)
2p
11
(4
11

22

2
12
)
=
4w
1

11

22
w
1

2
12
4p
1

11

22
+ p
1

2
12
2
11

12
w
2
+ 2p
2

11

12
+ w
1

2
12
p
1

2
12
2p
11
(4
11

22

2
12
)
=
4w
1

11

22
4p
1

11

22
2
11

12
w
2
+ 2p
2

11

12
2p
11
(4
11

22

2
12
)
=
2w
1

22
2p
1

22

12
w
2
+ p
2

12
p (4
11

22

2
12
)
=
2w
1

22

12
w
2
+ p (
2

12
2
1

22
)
p (4
11

22

2
12
)
(72)
6.3. Necessary and sufcient conditions for a maximum. Consider the Hessianof the prot equa-
tion.

2
(x
1
, x
2
) = p
_
2
11

12

12
2
22
_
=
_
2p
11
p
12
p
12
2p
22
_
(73)
For a maximumwe need the diagonal elements to be negative and the determinant to be positive.
The diagonal elements will negative if
11
and
22
are negative. The determinant of the Hessian is
given by

2p
11
p
12
p
12
2p
22

= 4p
2

11

22
p
2

2
12
= p
2
_
4
11

22

2
12
_
(74)
So the solution will be a maximum if
11
and
22
are both less than zero and 4
11

22
>
2
12
.
Notice that if 4
11

22
=
2
12
, the test fails.
6.4. Optimal output. We can obtain optimal output as a function of p, w
1
and w
2
by substituting
the optimal values of x
1
and x
2
into the production function.
18 PROFIT FUNCTIONS
y(p, w
1
, w
2
) = f(x
1
(p, w
2
, w
2
), x
2
(p, w
1
, w
2
))
=
1
_
2w
1

22
2p
1

22

12
w
2
+ p
2

12
p (4
11

22

2
12
)
_
+
2
_
2
11
w
2
2p
2

11
w
1

12
+ p
1

12
p (4
11

22

2
12
)
_
+
11
_
2w
1

22
2p
1

22

12
w
2
+ p
2

12
p (4
11

22

2
12
)
_
2
+
12
__
2w
1

22
2p
1

22

12
w
2
+ p
2

12
p (4
11

22

2
12
)
_ _
2
11
w
2
2p
2

11
w
1

12
+ p
1

12
p (4
11

22

2
12
)
__
+
22
_
2
11
w
2
2p
2

11
w
1

12
+ p
1

12
p (4
11

22

2
12
)
_
2
(75)
We can simplify equation 75 by simplifying the squared and cross product terms. First consider
x
2
1
.
x
2
1
=
_
2w
1

22
2p
1

22

12
w
2
+ p
2

12
p (4
11

22

12
)
2
_
_
2w
1

22
2p
1

22

12
w
2
+ p
2

12
p (4
11

22

2
12
)
_
=
_
4w
2
1

2
22
4pw
1

2
22
2w
1
w
2

12

22
+ 2pw
1

12

22
p
2
(4
11

22

2
12
)
2
_
+
_
4pw
1

2
22
+ 4p
2

2
1

2
22
+ 2pw
2

12

22
2p
2

12

22
p
2
(4
11

22

2
12
)
2
_
+
_
2w
1
w
2

12

22
+ 2p
1
w
2

12

22
+ w
2
2

2
12
pw
2

2
12
p
2
(4
11

22

2
12
)
2
_
+
_
2pw
1

12

22
2p
2

12

22
pw
2

2
12
+ p
2

2
2

2
12
p
2
(4
11

22

2
12
)
2
_
(76)
We can simplify equation 76 as follows
PROFIT FUNCTIONS 19
x
2
1
=
_
4w
2
1

2
22
8pw
1

2
22
4w
1
w
2

12

22
+ 4pw
1

12

22
p
2
(4
11

22

2
12
)
2
_
+
_
4p
2

2
1

2
22
+ 4pw
2

12

22
4p
2

12

22
p
2
(4
11

22

2
12
)
2
_
+
_
w
2
2

2
12
2pw
2

2
12
p
2
(4
11

22

2
12
)
2
_
+
_
p
2

2
2

2
12
p
2
(4
11

22

2
12
)
2
_
(77)
Simplifying again we obtain
x
2
1
=
_
4p
2

2
1

2
22
4p
2

12

22
+ p
2

2
2

2
12
p
2
(4
11

22

2
12
)
2
_
+
_
4pw
1

12

22
8pw
1

2
22
+ 4pw
2

12

22
2p
2
w
2

2
12
p
2
(4
11

22

2
12
)
2
_
+
_
4w
2
1

2
22
4w
1
w
2

12

22
+ w
2
2

2
12
p
2
(4
11

22

2
12
)
2
_
=
_
p
2
_
4
2
1

2
22
4
1

12

22
+
2
2

2
12
_
p
2
(4
11

22

2
12
)
2
_
+
_
4pw
1
_

12

22
2
1

2
22
_
p
2
(4
11

22

2
12
)
2
_
+
_
2pw
2
_
2
1

12

22

2

2
12
_
p
2
(4
11

22

2
12
)
2
_
+
_
4w
2
1

2
22
4w
1
w
2

12

22
+ w
2
2

2
12
p
2
(4
11

22

2
12
)
2
_
(78)
Similarly for x
2
20 PROFIT FUNCTIONS
x
2
2
=
_
4w
2
2

2
11
8pw
2

2
11
+ 4p
2

2
2

2
11
4w
1
w
2

11

12
p
2
(4
11

22

2
12
)
2
_
+
_
4p
1
w
2

11

12
+ 4pw
1

11

12
4p
2

11

12
p
2
(4
11

22

2
12
)
2
_
+
_
w
2
1

2
12
2pw
1

2
12
p
2
(4
11

22

2
12
)
2
_
+
_
p
2

2
1

2
12
p
2
(4
11

22

2
12
)
2
_
=
_
p
2
_
4
2
2

2
11
4
1

11

12
+
2
1

2
12
_
p
2
(4
11

22

2
12
)
2
_
+
_
2pw
1
_
2
2

11

12

1

2
12
_
p
2
(4
11

22

2
12
)
2
_
+
_
4pw
2
_
2
1

11

12
2
2

2
11
_
p
2
(4
11

22

2
11
)
2
_
+
_
w
2
1

2
12
4w
1
w
2

11

12
+ 4w
2
2

2
11
p
2
(4
11

22

2
12
)
2
_
(79)
And for x
1
x
2
x
1
x
2
=
_
2w
1

22

12
w
2
+ p (
2

12
2
1

22
)
p (4
11

22

2
12
)
_ _
2
11
w
2
w
1

12
+ p (
1

12
2
2

11
)
p (4
11

22

2
12
)
_
=
_
4w
1
w
2

11

22
2w
2
1

12

22
+ 2pw
1

22
(
1

12
2
2

11
)
p
2
(4
11

22

2
12
)
2
_
+
_
2w
2
2

11

12
+ w
1
w
2

2
12
pw
2

12
(
1

12
2
2

11
)
p
2
(4
11

22

2
12
)
2
_
+
_
2pw
2

11
(
2

12
2
1

22
) pw
1

12
(
2

12
2
1

22
) + p
2
(
1

12
2
2

11
) (
2

12
2
1

22
)
p
2
(4
11

22

2
12
)
_
(80)
Now collect terms in equation 80
PROFIT FUNCTIONS 21
x
1
x
2
=
_
p
2
(
1

12
2
2

11
) (
2

12
2
1

22
) + 2pw
1

22
(
1

12
2
2

11
) pw
2

12
(
1

12
2
2

11
)
p
2
(4
11

22

2
12
)
2
_
+
_
2pw
2

11
(
2

12
2
1

22
) pw
1

12
(
2

12
2
1

22
)
p
2
(4
11

22

2
12
)
2
_
+
_
2w
2
1

12

22
+ w
1
w
2

2
12
+ 4w
1
w
2

11

22
2w
2
2

11

12
p
2
(4
11

22

2
12
)
2
_
=
_
p
2
(
1

12
2
2

11
) (
2

12
2
1

22
)
p
2
(4
11

22

2
12
)
2
_
+
_
pw
1
(2
22
(
1

12
2
2

11
)
12
(
2

12
2
1

22
))
p
2
(4
11

22

2
12
)
2
_
+
_
pw
2
(2
11
(
2

12
2
1

22
)
12
(
1

12
2
1

11
))
p
2
(4
11

22

2
12
)
2
_
+
_
2w
2
1

12

22
+ w
1
w
2
_
4
11

22
+
2
12
_
2w
2
2

11

12
p
2
(4
11

22

2
12
)
2
_
(81)
Now substitute the results from equations 70, 72 78 , 79 and 81 into equation 75
22 PROFIT FUNCTIONS
y(p, w
1
, w
2
) = f(x
1
(p, w
2
, w
2
), x
2
(p, w
1
, w
2
))
=
1

2w
1

22

12
w
2
+ p (
2

12
2
1

22
)
p

4
11

22

2
12

+
2

2
11
w
2
w
1

12
+ p (
1

12
2
2

11
)
p

4
11

22

2
12

+
11

p
2

4
2
1

2
22
4
1

12

22
+
2
2

2
12

+ 4pw
1

12

22
2
1

2
22

p
2

4
11

22

2
12

+
11

2pw
2

2
1

12

22

2
12

+ 4w
2
1

2
22
4w
1
w
2

12

22
+ w
2
2

2
12
p
2

4
11

22

2
12

+
12

p
2
(
1

12
2
2

11
) (
2

12
2
1

22
)
p
2

4
11

22

2
12

2
+
pw
1
(2
22
(
1

12
2
2

11
)
12
(
2

12
2
1

22
))
p
2

4
11

22

2
12

+
12

pw
2
(2
11
(
2

12
2
1

22
)
12
(
1

12
2
1

11
))
p
2

4
11

22

2
12

2
+
2w
2
1

12

22
+ w
1
w
2

4
11

22
+
2
12

2w
2
2

11

12
p
2

4
11

22

2
12

+
22

p
2

4
2
2

2
11
4
1

11

12
+
2
1

2
12

p
2

4
11

22

2
12

2
+
2pw
1

2
2

11

12

2
12

p
2

4
11

22

2
12

+
22

4pw
2

2
1

11

12
2
2

2
11

p
2

4
11

22

2
11

2
+
w
2
1

2
12
4w
1
w
2

11

12
+ 4w
2
2

2
11
p
2

4
11

22

2
12

(82)
To start simplifying, we need to put everything over a common denominator
PROFIT FUNCTIONS 23
y(p, w
1
, w
2
) = f(x
1
(p, w
2
, w
2
), x
2
(p, w
1
, w
2
))
=

2pw
1

22

4
11

22

2
12

pw
2

12

4
11

22

2
12

+ p
2

4
11

22

2
12

(
2

12
2
1

22
)
p
2

4
11

22

2
12

2pw
2

11

4
11

22

2
12

pw
1

12

4
11

22

2
12

+ p
2

4
11

22

2
12

(
1

12
2
2

11
)
p
2

4
11

22

2
12

p
2

11

4
2
1

2
22
4
1

12

22
+
2
2

2
12

+ 4pw
1

11

12

22
2
1

2
22

p
2

4
11

22

2
12

2pw
2

11

2
1

12

22

2
12

+ 4w
2
1

11

2
22
4w
1
w
2

11

12

22
+ w
2
2

11

2
12
p
2

4
11

22

2
12

p
2

12
(
1

12
2
2

11
) (
2

12
2
1

22
)
p
2

4
11

22

2
12

2
+
pw
1

12
(2
22
(
1

12
2
2

11
)
12
(
2

12
2
1

22
))
p
2

4
11

22

2
12

pw
2

12
(2
11
(
2

12
2
1

22
)
12
(
1

12
2
1

11
))
p
2

4
11

22

2
12

2w
2
1

2
12

22
+ w
1
w
2

12

4
11

22
+
2
12

2w
2
2

11

2
12
p
2

4
11

22

2
12

p
2

22

4
2
2

2
11
4
1

11

12
+
2
1

2
12

p
2

4
11

22

2
12

2
+
2pw
1

22

2
2

11

12

2
12

p
2

4
11

22

2
12

4pw
2

22

2
1

11

12
2
2

2
11

p
2

4
11

22

2
11

2
+
w
2
1

2
12

22
4w
1
w
2

11

12

22
+ 4w
2
2

2
11

22
p
2

4
11

22

2
12

(83)
Notice that we have terms in the following p
2
, pw
1
pw
2
, w
2
1
, w
1
w
2
and w
2
2
. Now rearrange
terms in equation 83 combining like terms. After some manipulations we obtain
y(p, w
1
, w
2
) =
w
2
2

11
w
1
w
2

12
+ w
2
1

22
p
2
_

2
2

11

12
+
2
1

22
_
p
2
(4
11

22

2
12
)
(84)
6.5. The example prot function. We obtain the prot function by substituting the optimal values
of x
1
and x
2
into the prot equation
(p, w
1
, w
2
) = pf(x
1
(p, w
2
, w
2
), x
2
(p, w
1
, w
2
)) w
1
x
1
(p, w
2
, w
2
) w
2
x
2
(p, w
2
, w
2
)
= p
_
w
2
2

11
w
1
w
2

12
+ w
2
1

22
p
2
_

2
2

11

1

12
+
2
1

22
_
p
2
(4
11

22

2
12
)
_
w
1
_
2w
1

22

12
w
2
+ p (
2

12
2
1

22
)
p (4
11

22

2
12
)
_
w
2
_
2
11
w
2
w
1

12
+ p (
1

12
2
2

11
)
p (4
11

22

2
12
)
_
(85)
We can simplify equation 85 as follows
24 PROFIT FUNCTIONS
(p, w
1
, w
2
) =
_
w
2
2

11
w
1
w
2

12
+ w
2
1

22
p
2
_

2
2

11

1

12
+
2
1

22
_
p (4
11

22

2
12
)
_
w
1
_
2w
1

22

12
w
2
+ p (
2

12
2
1

22
)
p (4
11

22

2
12
)
_
w
2
_
2
11
w
2
w
1

12
+ p (
1

12
2
2

11
)
p (4
11

22

2
12
)
_
=
_
w
2
2

11
w
1
w
2

12
+ w
2
1

22
p
2
_

2
2

11

1

12
+
2
1

22
_
p (4
11

22

2
12
)
_
+
_
2w
2
1

22
+ w
1
w
2

12
pw
1
(
2

12
2
1

22
)
p (4
11

22

2
12
)
_
+
_
2w
2
2

11
+ w
1
w
2

12
pw
2
(
1

12
2
2

11
)
p (4
11

22

2
12
)
_
=
_
w
2
2

11
+ w
1
w
2

12
w
2
1

22
p
2
_

2
2

11

12
+
2
1

22
_
p (4
11

22

2
12
)
_
+
_
pw
1
(
2

12
2
1

22
)
p (4
11

22

2
12
)
_
+
_
pw
2
(
1

12
2
2

11
)
p (4
11

22

2
12
)
_
(86)
We can write equation 86 in the following useful fashion
(p, w
1
, w
2
) =

w
2
2

11
+ w
1
w
2

12
w
2
1

22
pw
1
(
2

12
2
1

22
) pw
2
(
1

12
2
2

11
) p
2

2
2

11

12
+
2
1

22

4
11

22

2
12

(87)
6.6. Optimal input demands. We can nd the optimal input demands by taking the derivative of
equation 86 with respect to w
1
and w
2
. First with respect to w
1
(p, w
1
, w
2
)
w
1
=
_

12
w
2
2
22
w
1
p (
2

12
2
1

22
)
p (4
11

22

2
12
)
_
x
1
=
_

12
w
2
+ 2
22
w
1
+ p (
2

12
2
1

22
)
p (4
11

22

2
12
)
_
(88)
Then with respect to w
2
(p, w
1
, w
2
)
w
2
=
_
2w
2

11
+ w
1

12
p (
1

12
2
2

11
)
p (4
11

22

2
12
)
_
x
2
=
_
2w
2

11
w
1

12
+ p (
1

12
2
2

11
)
p (4
11

22

2
12
)
_
(89)
6.7. Optimal output. Take the derivative of equation 87 with respect to p
PROFIT FUNCTIONS 25
(p, w
1
, w
2
) =

w
2
2

11
+ w
1
w
2

12
w
2
1

22
pw
1
(
2

12
2
1

22
) pw
2
(
1

12
2
2

11
) p
2

2
2

11

1

12
+
2
1

22

4
11

22

2
12

(p, w
1
, w
2
)
p
=

4
11

22

2
12

w
1
(
2

12
2
1

22
) w
2
(
1

12
2
2

11
) 2p

2
2

11

1

12
+
2
1

22

p
2

4
11

22

2
12

w
2
2

11
+ w
1
w
2

12
w
2
1

22
pw
1
(
2

12
2
1

22
) pw
2
(
1

12
2
2

11
) p
2

2
2

11

1

12
+
2
1

22

4
11

22

2
12

p
2

4
11

22

2
12

2
=

pw
1
(
2

12
2
1

22
) pw
2
(
1

12
2
2

11
) 2p
2

2
2

11

1

12
+
2
1

22

4
11

22

2
12

p
2

4
11

22

2
12

2
+

w
2
2

11
w
1
w
2

12
+ w
2
1

22
+ pw
1
(
2

12
2
1

22
) + pw
2
(
1

12
2
2

11
) + p
2

2
2

11

1

12
+
2
1

22

4
11

22

2
12

p
2

4
11

22

2
12

2
=

p
2

2
2

11

1

12
+
2
1

22

w
2
2

11
w
1
w
2

12
+ w
2
1

22

4
11

22

2
12

p
2

4
11

22

2
12

2
=

p
2

2
2

11

1

12
+
2
1

22

w
2
2

11
w
1
w
2

12
+ w
2
1

22

p
2

4
11

22

2
12

=
w
2
2

11
w
1
w
2

12
+ w
2
1

22
p
2

2
2

11

1

12
+
2
1

22

p
2

4
11

22

2
12

(90)
7. SINGULARITY OF THE HESSIAN MATRIX
We can showthat the Hessianmatrix of the prot function is singular. This is a direct implication
of linear homogeneity. To show this rst write the identity
(t p, tw) = t (p, w), where t is a scaler greater than zero (91)
Differentiate equation 91 with respect to t. This will yield
m

j=1

(t p
j
)
p
j
+
n

i=1

(t w
i
)
w
i
= (p, w) (92)
Differentiate equation 92 with respect to t again. Each term in equation 92 will yield a sum of
derivatives so the result will be a double sum as follows
m

j=1
m

k=1

(t p
k
) (t p
j
)
p
k
p
j
+
m

j=1
n

=1

(t w

) (t p
j
)
w

p
j
+
n

i=1
m

k=1

(t p
k
) (t w
i
)
p
k
w
i
+
n

i=1
n

=1

(t w

) (t w
i
)
w

w
i
= 0
(93)
We can factor t out of each terms in equation 93 because
f(x)
(t x)
=
1
t
f(x)
x
(94)
This will give
26 PROFIT FUNCTIONS
1
t
2
m

j=1
m

k=1

p
k
p
j
p
k
p
j
+
1
t
2
m

j=1
n

=1

p
j
w

p
j
+
1
t
2
n

i=1
m

k=1

p
k
w
i
p
k
w
i
+
1
t
2
n

i=1
n

=1

w
i
w

w
i
= 0
(95)
This then implies that
m

j=1
m

k=1

p
k
p
j
p
k
p
j
+
m

j=1
n

=1

p
j
w

p
j
+
n

i=1
m

k=1

p
k
w
i
p
k
w
i
+
n

i=1
n

=1

w
i
w

w
i
= 0
(96)
The last expression in equation 96 is actually a quadratic form involving the Hessian of (p,w).
To see this write out the Hessian and then pre and post multiply by a vector containing output and
input prices.

p1 p2 . . . pm w1 . . . wn

p
2
1

p
2
p
1
. . .

2

pmp
1

w
1
p
1
. . .

2

wnp
1

p
1
p
2

p
2
2
. . .

2

pmp
2

w
1
p
2
. . .

2

wnp
2
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

p
1
pm

p
2
pm
. . .

2

p
2
m

w
1
pm
. . .

2

wnpm

p
1
w
1

p
2
w
1
. . .

2

pmw
1

w
2
1
. . .

2

wnw
1
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

p
1
wn

p
2
wn
. . .

2

pmwn

w
1
wn
. . .

2

w
2
n

p1
p2
.
.
.
pn
w1
w2
.
.
.
wn

(97)
Equation 96 implies that this quadratic form is zero for this particular vector (p,w). This means
that the Hessian matrix is singular. The prot function is convex so we know that its Hessian
is positive semi-denite. This implies that it is positive semidenite only and cannot be positive
denite.
8. SENSITIVITY ANALYSIS FOR THE PROFIT MAXIMIZATION PROBLEM
The Hessian of the prot function is positive semidenite so all its diagonal elements are non-
negative (Hadley [9, p. 117]).
8.1. Response of Output Supply. Consider rst the response of any output to its own price.
y
j
p
j
=

2

p
2
j
0 (98)
Supply curves derived from prot maximization will always slope upwards because this deriv-
ative is on the diagonal of the Hessian. Now consider the response of an output to another product
price.
y
j
p
k
=

2

p
k
p
j
(99)
PROFIT FUNCTIONS 27
This element is not on the diagonal of the Hessian of the prot function and so we cannot deter-
mine its sign.
Similarly we cannot sign the response of an output to an input price.
y
j
w
i
=

2

w
i
p
j
(100)
8.2. Response of Input Supply. Consider rst the response of any input to its own price.
x
i
w
i
=

2

w
2
i
0 (101)
Demand curves derived from prot maximization will always slope downwards because this
derivative is on the diagonal of the Hessian and input demand is the negative of the derivative of
the prot function with respect to input price. Now consider the response of an input to another
input price.
x
i
w

=

2

w
i
(102)
This element is not on the diagonal of the Hessian of the prot function and so we cannot deter-
mine its sign.
Similarly we cannot sign the response of an input to an output price.
x
i
p
k
=

2

p
k
w
i
(103)
8.3. Homogeneity. Because output supply and input demand are derivatives of a function which
is homogeneous of degree one, they are homogeneous of degree zero.
y
j
(p, w) = y
j
(p, w)x
i
(p, w) = x
i
(p, w) (104)
8.4. Symmetry of response. If (p, w) is twice differentiable then Youngs theoremon second cross
partial derivatives implies

p
i
p
j
=

2

p
j
p
i

p
k
w
i
=

2

w
i
p
k

w
i
w
j
=

2

w
j
w
i
(105)
This then implies
y
j
p
i
=
y
i
p
j

w
i
p
k
=
y
k
w
i
x
i
w
j
=
x
j
w
i
(106)
28 PROFIT FUNCTIONS
The response of any output to a different output price is symmetric, i.e. cross price derivatives
are equal. Similarly with input demand response. The way an output responds to a particular
input price is the same as the response of that input to the outputs price.
8.5. Own and cross price response. Because any principal submatrix of a positive semi-denite
matrix is also positive semidenite (Hadley [9, p. 262]), we also have the following results.
x
i
w
i

x
j
w
j

x
i
w
j

x
j
w
i
(107)
To see this note that if the submatrix above is positive semi-denite then the determinant must
be non-negative which means that (the product of the diagonal elements) - (the product of the off
diagonal elements) must be greater than or equal to zero.
We also have
y
j
p
j

y
k
p
k

y
j
p
k

y
k
p
j
(108)
The product of own price responses is larger than the product in cross price responses.
We also have

y
j
p
j

x
k
w
k

y
j
w
k

x
k
p
j
(109)
9. SUBSTITUTES AND COMPLEMENTS
9.1. Gross substitutes. We say that inputs are gross substitutes if
x
i
(p, w)
w
j
=
x
j
(p, w)
w
i
0 (110)
9.2. Gross complements. We say that inputs are gross complements if
x
i
(p, w)
w
j
=
x
j
(p, w)
w
i
0 (111)
10. RECOVERING PRODUCTION RELATIONSHIPS FROM THE PROFIT FUNCTION
10.1. Marginal Products. We can nd marginal products from the rst order conditions for prot
maximization. The easiest way to see this is when technology is represented by an asymmetric
transformation function.
= max
x y
[p
1
f( y, x) +
m

2=1
p
j
y
j

n

i=1
w
i
x
i
] (112a)
= max
x y
[p
1
f( y, x) + p y wx] where
p = (p
2
, p
3
, . . . , p
m
), y = (y
2
, y
3
, . . . , y
m
),
w = (w
1
, w
2
, . . . , w
n
) x = (x
1
, x
2
, . . . , p
n
)
If f( y,x) is differentiable, then the rst order conditions for maximizing prot are as follows.
PROFIT FUNCTIONS 29

x
i
= p
1
f ( y, x)
x
i
w
i
= 0, i = 1, 2, . . . , n

y
j
= p
1
f ( y, x)
y
j
+ p
j
= 0, j = 2, 3, . . . , m
(113)
We can then obtain marginal products by rearranging the rst order conditions. We can nd the
impact on y
1
of an increase in any of the inputs as
f ( y, x)
x
i
=
w
i
p
1
, i = 1, 2, . . . , n (114)
We can nd the impact on y
1
of an increase in any of the other outputs as
f ( y, x)
y
j
=
p
j
p
1
, j = 2, 3, . . . , m (115)
We can nd other marginal products by solving the prot maximizationproblem witha different
normalization than the one on y
1
.
In the case of a single output we obtain
f (x)
x
i
=
w
i
p
, i = 1, 2, . . . , n (116)
This is clear from gure 2. The isoprot line is given by
= py wx
Solving for y we obtain
y =

p
+
w
p
x
FIGURE 2. Marginal Product is Equal to Price Ratio
x
y
fx
30 PROFIT FUNCTIONS
10.2. Output elasticity. We normally think of output elasticity in terms of one output and many
inputs so dene it as follows.

i
=
f(x)
x
i
x
i
y
(117)
If we substitute for
f(x)
xi
from equation 116 we obtain

i
=
w
i
p
x
i
y
(118)
10.3. Elasticity of scale. Elasticity of scale is given by the sum of the output elasticities,i.e.,
=
n

i=1
f(p, w)
x
i
x
i
y
=
n

i=1
w
i
p
x
i
y
=

n
i=1
w
i
x
i
p y
=
Cost
Revenue
(119)
A prot maximizing rm will never produce in a region of the production function with in-
creasing returns to scale assuming the technology eventually becomes locally concave and remains
concave. As the result for a prot maximizing rm, returns to scale is just the ratio of cost to rev-
enue.
11. DUALITY BETWEEN THE PROFIT FUNCTION AND THE TECHNOLOGY
We can showthat if we have an arbitrarily specied prot function that satises the conditions in
section 3, we can construct fromit a reasonable technology that satises the conditions we specied
in section 1.2.4 and that would generate the specied prot function. A general discussion of this
duality is contained in Diewert [2]
11.1. Constructing a technology from the prot function. Let T* be dened as
T

{(x, y) : py wx (p, w), (p, w) (0, 0)} (120)

To see the intuition of why we can construct T this way consider the case of one input and one
output as in gure 3
If we pick a particular set of prices then T

consists of all points below the line that is tangent

to T(x,y) at the optimal input output combination. The equation {(x, y) : py wx (p, w) for
a particular set of prices denes a line in R
2
or a hyperplane in R
n+m
. Points below the line are
considered to be in T

. Now pick a different set of prices with w lower and construct a different
line as in gure 4. The number of points in T

is now less than before. If we add a third set of

prices with w higher than in the original case as in gure 5, we can reduce the size of T

even more.
Continuing in this fashion we can recover T(x,y). This is an application of what is often called
Minkowskis theorem.
To show that T is in fact equal to T

, we need to show that T

T. We do this by assuming that

some particular input output combination (x
0
, y
0
) T

but (x
0
, y
0
) / T. If (x
0
, y
0
) / T, then (x
0
,
PROFIT FUNCTIONS 31
FIGURE 3. Half-Spaces and Technology Set
x
y
Tx,y
FIGURE 4. Two Half-Spaces and Technology Set
x
y
Tx,y
y
0
) can be separated from T by a hyperplane (py
0
- wx
0
). But T

consists of points lying below the

hyperplane dened by the maximal prots with prices p and w. This means that the point (x
0
, y
0
)
cannot be in T

. We can show this in gure 6. Consider point a in gure 6 which is not in T(x,y). It
is in T

if we only consider the half spaces dened by the hyperplanes intersecting T(x,y) at points
c and d. But at the prices dened by a hyperplane with slope as the one passing through b, point
a is not in T

. If we consider any other point not in T, there exists a hyperplane that also excludes
it from T

. This implies than that T

T and so T

T. For more on this topic consult McFadden

[12] or Fare and Primont [6].
11.2. Using the properties of the prot function to recover properties of the technology. We
repeat here for convenience the properties of T(x,y) and (p,w).
11.2.1. Properties of the technology.
32 PROFIT FUNCTIONS
FIGURE 5. Three Half-Spaces and Technology Set
x
y
Tx,y
FIGURE 6. Separating Hyperplanes and the Technology Set
x
y
Tx,y
a
b
c
d
T.1a: Inaction and No Free Lunch. (0,y) T x R
n
+
and y R
m
+
. This implies that T(x,y) is a
non-empty subset of R
m+n
+
T.1b: (0,y) T, y 0, y = 0.
T.2: Input Disposability. If (x,y) T and 1 then (x, y) T.
T.2.S: Strong Input Disposability. If (x,y) T and x x, then (x, y) T.
T.3: Output Disposability. (x,y) R
n+m
+
,if (x, y) T and 0 < 1 then (x, y) T.
T.3.S: Strong Output Disposability. If (x,y) T and y y, then (x, y) T.
T.4: Boundedness. For every nite input vector x 0, the set y P(x) is bounded from above.
This implies that only nite amounts of output can be produced from nite amounts of
inputs.
T.5: T(x) is a closed set. The assumption that P(x) and V(y) are closed does not imply that T
is a closed set, so it is assumed. Specically, if [x

x
0
, y

y
0
) and (x

, y

) T, ] then
(x
0
y
0
) T.
PROFIT FUNCTIONS 33
T.9: T is a convex set. This is not implied by the convexity of P(x) and V(y). Specically, a
technology could exhibit a diminishing rate of technical substitution, a diminishing rate of
product transformation and still exhibit increasing returns to scale.
11.2.2. Properties of the prot function.
.1: (p, w) is an extended real valued function (it can take on the value of + for nite
prices) dened for all (p, w) (0
m
, 0
n
)) and (p, w) pa - wb for a xed vector (a, b)
(0
m
, 0
n
). This implies that (p, w) 0 if (0
m
, 0
n
) T(x,y),which we normally assume.
.2: is nonincreasing in w
.3: is nondecreasing in p
.4: is a convex function
.5: is homogeneous of degree 1 in p and w.
We will consider only a few of the properties of T(x,y)
11.2.3. T(x,y) is non-empty. By .1, (p, w) pa - wb for a xed vector (a, b) (0
m
, 0
n
) or (p, w)
0 if (0
m
, 0
n
) T(x,y). Given that (p, w) 0, with x zero, there are obviously values of y (for
example 0) which make (py - wx) less than (p, w). So T

(x,y) is not empty and (0,y) T

(x,y).
11.2.4. Strong Input Disposability. If (x,y) T and x x, then (x, y) T. .
Suppose that (x, y) T

, 0 x x and (p, w) 0. Then py - wx py - wx (p, w) (p, w)

0. Thus (y, x) T

.
11.2.5. Strong Output Disposability. If (x,y) T and y y, then (x, y) T. .
Suppose (y, x) T

and 0 y y then py - wx py - wx (p, w) (p, w) 0. Thus (y, x)

T

.
11.2.6. T(x) is a closed and convex set. From the denition, T* is the intersection of a family of closed
half-spaces because we have a hyperplane in n+m space dividing the space. Thus T is a closed
convex set (Rockafellar [13, p. 10].
34 PROFIT FUNCTIONS
REFERENCES
[1] Chambers, R. G. Applied Production Analysis. Cambridge: Cambridge University Press, 1988.
[2] Diewert, W. E. Functional Forms for Prot and Transformation Functions. Journal of Economic Theory.6(1973):284-316.
[3] Diewert, W. E. Applications of Duality Theory in Frontiers of Quantitative Economics, Vol. 2, ed. M. Intriligator and D.
Kendrick. Amsterdam: North Holland, 1974.
[4] Diewert, W.E. Duality Approaches to Microeconomic Theory in Handbook of Mathematical Economics, Vol 2, ed. K.J.
Arrow and M.D. Intriligator. Amsterdam: North Holland, 1982
[5] Fare, R. Fundamentals of Production Theory. NewYork: Springer-Verlag, 1988.
[6] Fare, R. and D. Primont. Multi-output Production and Duality: Theory and Applications. Boston: Kluwer Academic Pub-
lishers, 1995.
[7] Ferguson, C. E. The Neoclassical Theory of Production and Distribution. Cambridge: Cambridge University Press, 1971.
[8] Fuss, M. and D. McFadden. Production Economics: A Dual Approach to Theory and Application. Amsterdam: North Hol-
land, 1978.
[9] Hadley, G. Linear Algebra. Reading, MA: Addison-Wesley, 1961.
[10] Lau, L. J. Characterization of the Normalized Restricted Prot Function Journal of Economic Theory12(Feb. 1976):131-
163.
[11] Lau, L. J. Applications of Prot Functions in Production Economics: A Dual Approach to Theory and Applications, ed. M.
Fuss and D. McFadden, Amsterdam: North-Holland, 1978.
[12] McFadden, D. Duality of Production, Cost and Prot Functions in Production Economics: A Dual Approach to Theory
and Applications, ed. M. Fuss and D. McFadden, Amsterdam: North-Holland, 1978.
[13] Rockafellar, R. T. Convex Analysis. Princeton New Jersey: Princeton University Press, 1970.
[14] Shephard, R. W. Theory of Cost and Production Functions. Princeton New Jersey: Princeton University Press, 1970.
[15] Varian, H. R. Microeconomic Analysis. 3
rd
Edition. NewYork: Norton, 1992.