Cost Function
Cost Function
Cost Function
Cost Minimization
Economists assume that firms minimize costs (this is consistent with profit
maximization). The math behind this goes back to the Lagrangian that we used in
looking at consumer decisions. In fact, a lot of producer theory is basically the same as
consumer theory.
Whereas a consumer sought to maximize utility given a budget constraint, a producer will
minimize costs given some quantity constraint. That is, if a producer will produce a
certain quantity of output, q0, they will seek to do so at the lowest possible cost.
The Lagrangian for this is:
Minimize wl + vk subject to f(k,l) = q0
L = wl + vk + (q0 f(k,l))
Taking derivatives with respect to l, k and and setting these equal to zero gives us:
L
= w f l = 0
l
L
= v f k = 0
k
L
= q 0 f (k , l ) = 0
w = f l
v f k = 0
v = f k
f
w f l
=
= l
v f k f k
So, just as with the consumers decision we had, for two goods, x and y, the relationship
that the price ratio equaled the ratio of the marginal utilities or the marginal rate of
substitution:
p x MU x
=
= MRS x , y
p y MU y
So with the producers cost minimizing input decision we have the relationship that the
ratio of the costs of capital and labor is equal to the ratio of their marginal products,
which is also known as the marginal rate of technical substitution (MRTS):
MPl
w fl
=
=
= MRTS l,k
v f k MPk
We can also get the relationship that
w v
=
=
fl fk
With the interpretation being that the marginal cost of increasing total output either by
adding labor or by adding capital is equal to , and this is the marginal cost in, for
example, dollars per unit.
A graph showing the cost minimizing point for the production function
q=kl.
The math underlying this, from the point of view of minimizing the cost of producing a
quantity of 120, is:
q = f (k , l ) = kl
f
MPl =
= fl = k
l
f
= fk = l
MPk =
k
MPl w
=
MPk
v
k 40
=
l 12
10
k= l
3
q = 120 kl = 120
10
l l = 120
3
l 2 = 36
l=6
10
10
k = l = 6 = 20
3
3
Expansion Paths
Expansion paths answer the question, How does the combination of capital and labor
change as the size of a firm increases? That is, as you move to higher and higher
isoquants, how does the cost-minimizing bundle of capital and labor change, holding w
and v constant?
Expansion paths are shown in Figures 8.2 and 8.3 in the textbook.
If an expansion path is a straight line, the relative mix of capital and labor remains
constant as the quantity produced increases. So, the same proportions of capital and labor
would be used. If a production function is homothetic, its expansion path will be a
straight line. Both the Cobb-Douglas and CES production functions are homothetic. In
some sense this means that the optimal mix of labor and capital doesnt actually depend
on the quantity of output produced.
The fact that a Cobb-Douglas production function is homothetic is demonstrated by
equation 8.10, which shows that the ration of capital to labor is independent of the total
quantity of the good that is being produced and is simply a function of the ratio of w to v:
k w
=
l
v
(8.10)
Cost Functions
Cost functions assume that a firm is cost minimizing, that is that it will produce any
quantity of output at the lowest possible cost.
A cost function expresses a firms total cost as a function of the cost of labor, w, the cost
of capital, v, and the total quantity of output to be produced, q:
C = C(v, w, q)
A cost function is increasing (or at least non-decreasing) in all of these arguments. That
is, as v, w or q increase, the cost of production will not fall and will probably rise.
Average cost is defined as:
AC(v, w , q ) =
C(v, w , q )
q
Marginal cost, or the additional cost incurred to produce an additional unit, is defined as:
MC(v, w , q ) =
C(v, w , q )
q
Marginal cost is the partial derivative of total cost with respect to quantity. It is also the
slope of the total cost function:
A graph showing an upward sloping total cost curve and showing that the
slope of the total cost curve is the marginal cost.
For example, if C = 50 + wvq2, then the average and marginal cost functions would be:
AC =
C 50
=
+ wvq
q
q
MC =
C
= 2 wvq
q
Two graphs, the first showing an upward sloping total cost curve and the
second showing a marginal cost and an average cost curve, with the
marginal cost curve crossing the average cost curve at its minimum.
As an example, consider the cost function C(q) = 420 + 2q + q2. You should be able to
plot out the total cost, average cost and marginal cost functions to confirm that they look
like the curves shown above.
Second, this could be drawn with no fixed costs and constant marginal cost. This would
be the case, for example, if there were no fixed costs and the cost of production were
constant at $5 per unit. Then both the marginal cost and the average cost would be
constant. This is shown below.
A graph showing an upward sloping, linear total cost curve and the
associated horizontal marginal cost and average cost curves.
As an example of this type of cost function, consider C(q) = 5q. The marginal cost
function is just MC(q)=5 and the average cost function is AC(q)=5.
q
= 0.6k 0.3 l 0.4
l
q
MPk =
= 0.3k 0.7 l 0.6
k
MPl =
2kv
lw
,k =
w
2v
q = k 0.3 l 0.6
2kv
q=k
w
qw 0.6
0.9
k = 0.6 0.6
2 v
0.6
0.3
qw 0.6
k* = 0.6 0.6
2 v
0.9
0.3
l 0.9 w 0.3
lw
q = l 0.6 = 0.3 0.3
2 v
2v
0.3 0.3
q2 v
l 0.9 =
w 0.3
1
l* =
0.3
w
Plugging these functions for k* and l* into the definition of total costs gives us:
C(v, w , q ) = w l * + v k *
1
0.6 0.6
+
v
C(v, w , q ) = w
0.3
2 v
1
2 1
2
1 1 2
C(v, w , q ) = q 0.9 2 3 v 3 w 3 + 2 3 v 3 w 3
0.9
1
1
2
2
1
C(v, w , q ) = q 0.9 v 3 w 3 2 3 + 2 3
2. Non-decreasing in q, v and w
If the quantity produced increases, or if the costs of capital or labor increase, production
wont become less expensive. This should be pretty clear.
Input Substitution
Dont worry about this too much. The two important things to remember are:
1. Firms can and will substitute one input for another as inputs become relatively more
or less expensive. So, if labor costs are currently 60% of a firms costs and there is the
potential for labor costs to double, this does not mean that a firms costs will rise by 60%
as they will likely substitute capital for labor in the production process.
2. The degree to which a firm may substitute one input for another varies from firm to
firm and from time to time. Some industries have great potential for the substitution of
one input for another while others, by their nature, do not. The assembly of automobiles
or digging of holes offers great opportunities for substitution of inputs while surgery or
massage, for the most part, do not.
inputs. This progress may be illustrated through changed in the production function or
through changes in the cost function.
One way to model technical change is to describe it as a function of time, t, basically
taking the assumption that technical change is a disembodied process that occurs at some
exogenous rate. While this isnt a reasonable real-world story, it might be a good and
simple analytical model.
In terms of a production function, the factor A(t) might represent a multiplier that shows
how much more output the same quantity of input can produce at time t versus some
initial time period, perhaps called period 0.
To put this in terms of a production function, we might say that the production function at
time t is given by:
q t = A( t )k l
If A(0)=1.0 and A(10)=1.5, this suggests that as time moves from period 0 to period 1,
technology improves so that the same set of inputs can produce 50% more output.
To put this in terms of the cost function (which is what the book does) we might state this
as:
C t (v, w , q ) =
C 0 (v, w , q )
A( t )
Where C0(v,w,q) is the cost function at time zero. If, as above, A(0)=1.0 and A(10)=1.5,
then costs would have fallen by 33% over this time period, a statement equivalent to
saying that productivity rose by 50%.
To think about the structure of the process behind A(t), you might imagine that some
neutral (that is, not favoring one input or another) technical progress occurs at some rate
of i% annually. This is the process described in the latter part of Example 8.3 where the
assumed rate of technical progress is 3% and the related production and cost functions
are:
q t = A( t )k 0.5 l 0.5 = e 0.03t k 0.5 l 0.5
C t (v, w , q ) =
C 0 (v, w , q )
e 0.03t
The term e0.03t is an expression for the continuous rate of change at 3% over time. The
number e is the irrational number from natural logarithms that is approximately equal to
2.71828 and for which the natural log is 1. That is, ln e = 1. So, holding inputs constant,
output could increase by 3% annually. Holding output constant, costs could fall by 3%
annually.
C
v
l c (v, w , q ) =
C
w
The underlying assumption is that the firm is minimizing its costs. The demand functions
are contingent (as indicated by the superscript c) on the quantity of output produced.
You should be sure you can work through this calculation for the Cobb-Douglas cost
function as given in Example 8.4
SC = vk1 + wl
In the short run, the firm is constrained in its choice of capital and can only vary its
output by varying the amount of labor it hires. As a result, a firm wont, in general. Be
minimizing the cost of producing the quantity of output that it is producing. This is
shown in Figure 8.7 where the firm is constrained to have capital of level k1. In this
figure, if they produce q0 or q2, they will produce at a cost higher than the minimum
possible cost for those quantities. Only q1 is consistent with cost minimization with
capital level k1.
Marginal cost, in the short run, is the marginal cost of additional output that is produced
through the hiring of additional labor. The book gives this as:
SMC =
Cost
SC
=
Output
q
(8.54)
Another way to think about short run marginal cost is that it is equal to the wage paid to
an extra unit of labor divided by the additional output that labor provides:
SMC =
Cost
wage
=
Output MPlabor
A graph showing that the marginal cost curves intersects the short run
average cost curve at the minimum of the short run average cost curve.
Note that average cost is U-shaped and the marginal cost curve intersects the average cost
curve at minimum average cost.
Second, you could diagram two short run average cost curves, each associated with a
different quantity of capital, on the same graph:
Two short run average cost curves, based on different levels of capital,
with minima at different quantities.
SRAC1 is associated with a lower level of capital and has a minimum at a lower quantity.
SRAC2 is associated with a higher level of capital and has a minimum at a higher
quantity.
Now, we could draw several short run average cost curves on the same graph. Again,
each would be associated with a different level of capital. To point out a typo, the curves
should be labeled SRAC and not SRAS, which is a bit of a slip back to teaching
macroeconomics for me.1
This is, in fact, an economics joke. If you thought it was funny, you may consider yourself an economist
with a decent sense of humor.
A series of four short run average cost curves and a U-shaped long run average cost
curve that runs along the bottom of the short run curves and is tangent to each of them.
The long run average cost curve is the average cost curve that is relevant if all inputs can
be altered. This will allow any particular quantity of output to be produced at the lowest
average cost possible for that quantity.
Practice Problems
1. For the production function q = k0.2l0.3, use a Lagrangian to solve for the costminimizing input bundle to produce 100 units of output if v=10 and w=15. What is the
marginal cost of production at the solution you find?
2.
a.
b.
c.
Find the average and marginal cost functions for each of the following cost functions.
v w
3. C( v, w , q ) = q +
a b
1
4. C( v, w , q) = q Bv w
5. C(q) = 49 + 5q + 3q2
Do the following questions from the textbook: 8.3, 8.5, 8.9a, 8.10a