Simon Fraser University Prof.

Karaivanov
Department of Economics Econ 201

COST MINIMIZATION

Pro t Maximization and Cost Minimization

Remember that the rm's problem is maximizing pro ts by choosing the optimal quantities
of inputs to employ and output to produce. We already know how to solve the rm's pro t
maximization problem in a competitive market environment. This direct approach is great but there
are some very important insights especially on the cost side of the rm's problem that can be
gained if we look at the rm's problem through a more indirect approach. The idea is to break down
the pro t maximization problem of the rm into two stages: rst, we look at the problem of minimizing
costs of producing any given amount of output, and second, we look at how to choose the most pro
table output level.
These notes cover the rst of these two stages, i.e. they answer the question: for a given
amount of output, y (just some number, 10, 20, 100, which does not have to be necessarily the
rm's optimal output that it will end up producing) what is the minimum cost of producing it. In
other words, we ask given that we have to produce y units of output and given input prices
w1 and w2, how much of each input should the rm employ to produce y in the least costly
way. Notice that the amount of each input that will be used will depend on the output level that the
rm wants to produce - if you want to produce 1000 units of outputs you would de nitely need more
workers and machines than to produce just 10 units.
Notice that cost minimization is a necessary condition for pro t maximization in competitive
markets. If, for a given level of output, one is not cost minimizing that means that he is also not pro t
maximizing. Why? For a given y revenue is xed (taking p as given), so if there is a less costly way
to produce this output level this will lead to higher pro ts.

The Cost Minimization Problem

Given the above discussion the rm wants to solve:
min
x ;x
w1x 1 + w2x 2
1 2

s:t: f(x1; x2) = y

How is the above interpreted? The rm wants to minimize its costs (w1x1 + w2x2) of
producing y units of output. The fact that the rm wants produce y units of output is given by the
constrain f(x1; x2) = y: Remember that the production function, f(x1; x2) corresponds to the maximum
output that can be extracted from x1 units of input 1 and x2 units of input
2 - i.e. (since inputs are costly), using the production function we would use x1 and x2 most e ciently.
The cost minimization is then done by choosing how much of each input to employ (x1 and x2) such
that the costs of producing y using the production function are minimized.

1
 How Do We Solve The Cost Minimization Problem?
In general the cost minimization problem is harder to solve that the consumer problem or
the pro t maximization problem since usually the constraint (f(x1; x2) = y) is a non-linear
function of x1 and x2 so sometimes it may be hard to express x2 in terms of x1 from it and plug into
the costs that we are minimizing. When this is the case there are two ways to proceed of which the
rst is simpler and recommended and the second is optional for those of you with more
knowledge of math (you can use either one at the exams):

1. Solving the cost minimization problem using the optimality condition T RS = w1 : w2

Remember that we showed in class graphically that at the optimal choice of input quantities (the
optimum) we must have that the isocost line is tangent to the isoquant corresponding to output level y:
The slope of the isoquant is given by the technical rate of substitution (TRS) which is a function of x1
and x2: Why? Because remember what the TRS is equal to: it equals
minus the ratio of marginal products, i.e. T RS = MP1(x1;xMP2)
2(x1;x2)
but the marginal products are

functions of x1 and x2 since they are simply the partial derivatives of the production function which is
a function of x1 and x2: The slope of the isocost line (remember from class) was w1 i.e. minus thew2
ratio
of the input prices (this is similar to the slope of the budget line, however note that the isocost curve
does not serve the same purpose as the budget line). In the consumer problem we had a xed budget
line and we were trying to nd the point on this budget line which would give us maximum utility - i.e.
the point which lies on the highest indi erence curve. In contrast, in the rm's problem we have a xed
isoquant and we are trying to nd the point on this xed isoquant which gives us minimum costs,
i.e. the point lying on the lowest isocost line.
Given the above, we can solve the cost minimization problem by solving two equations for
the two unknowns x1 and x2: the optimality condition and the constraint that y must be produced:
MP1(x1; x2) w1
= (optimality condition)
MP2(x1; x2) w2

f(x1; x2) = y (output must be y condition)

Notice that y is not an unknown! It is just a xed level of output we want to produce (say
10, 20, 0 or whatever). We do not have to solve for y - it is given.
The solutions to the above problem x^1 and x^2 will be generally functions of everything that is
given: the input prices, w1 and w2 and the output level y; i.e. we will have:
x^1 = x^1(w1; w2; y)
x^2 = x^2(w1; w2; y)

These solutions (to the cost minimization problem of a rm facing input prices w1 and w2
which wants to produce y) are called the conditional input demands for producing y units of
output at input rises w1 and w2: Why are they called conditional? Because they are conditioned (i.e.
depend on) what level of output we want to produce - if we want to produce more output more inputs
will be needed in general.

2
 Di erence between the conditional input demands from the cost minimization problem and
the (unconditional) input demands from the pro t maximization prob- lem
It is important to understand that the conditional input demands coming from the cost
minimization problem above are not the same thing as the (unconditional, as sometimes called) input
demands resulting from the pro t maximization problem (see the notes on pro t maximization
and input demands). Remember the pro t maximizing input demands,
x 1(w1; w2; p) and x2(w1; w2; p) were functions of the input prices, w1 and w2 and the output
price p: They are called unconditional (or simply not called conditional) because they do not depend
on y - the output level in the rm's pro t maximization problem is already optimized and this optimized
output level is a function of the prices (w1; w2; p) that are taken as given by the rm - thus the
(unconditional) input demands are only functions of what is given, i.e. the prices. In contrast, the cost
minimization problem is being solved for a given output level which is not necessarily the optimized
one that the rm will eventually choose. Why? Remember the cost minimization problem was only
the rst step in the indirect approach to solve the rm's problem and at that rst step the output level was
taken as given as we were only interested in how to produce any given output level at minimum cost.
Thus the conditional input demands depend only on what is given in the cost minimization problem -
i.e. the input prices, w1 and w2 and the output level y: One more time: the (unconditional) input
demands depend on output price, p, the conditional input demands depend on output level, y: Both of
them depend on input prices, w1 and w2:
2. Solving the cost minimization problem using the Lagrange method (OPTIONAL! NOT
REQUIRED FOR THE EXAM!)
For those of you interested in math and knowing what the Lagrange method of solving
optimization problems is, the cost minimization problem of the rm stated above can be solved also
using this method. What we do is write the Lagrangean:
(x1; x2; ) = w1x1 + w2x2 (f(x1; x2) y)
where is the Lagrange multiplier on the constraint. Then we have to take the partial deriva- tives of
the Lagrangean with respect to its three arguments, x1; x2 and and set them equal to zero:
@f(x1; x2)
w1 = 0
@x1
@f(x1; x2)
w2 = 0
@x2

f(x1; x2) y = 0
@f(x1;x2) @f(x1;x2)
where
@x1
and @x2
are the rst partial derivatives of the production function with

respect to x1 and x2: The last equation is simply our constraint. We can take the rst two
equations above to get:
@f(x1; x2)
w1 =
@x
1

3
 @f(x1; x2)
w2 = @x
2

Divide them through (both sides) to get:

@f(x1;x2)
w1 @x1
= @f(x1;x2)
w2 @x2

But what are the partial derivatives of the production - remember they are simply the marginal
products of each input, MP1(x1; x2) and MP2(x1; x2): Thus we get:
w1 MP1(x1; x2)
=
w2 MP2(x1; x2)
which is what we had before! Then this equation, together with the constraint f(x1; x2) = y
can be used to solve for the conditional input demands x^1(w1; w2; y) and x^2(w1; w2; y):

An Example With A Speci c Production Function:

1=3 1=3
Take f(x1; x2) = x x and let input prices be w1 = 1; w2 = 2: We want to solve the
1 2
rm's cost minimization problem of producing y units of output. We will use all three methods
discussed above and obtain the same results:
{ substitution from the constraint
{ using the optimality condition MP1(xMP
1;x2)
= w 1 w2
2(x1;x2)
{ using the Lagrange method (optional)
Let us rst write down the problem:
min 1x1 + 2x2
x1;x2

s.t. x1=3 1=3

1 x2 =y
Method 1 (Direct Substitution)

1. We rst have to express x2 in terms of x1 from the technology constraint f(x1; x2) = y: The easiest
way to to it is get rid of the exponents rst by raising the constraint to power 3 - we get:
x1 x 2 = y 3
3
y
thus x2 = :
x1

2. Now let's plug the expression for x2 obtained above in our cost function to get the
following minimization problem:
y3
min x 1 + 2
x1 x1
3. We solve the above problem of just one unknown (x1) by taking the rst derivative of
the function being minimized and setting it to zero:
1
4
1 + 2y3( )=0
x 21

5
 1
(we use that the derivative of 1 is ): Remember y is given, i.e. treated as if it were a
x1 x21

number!. From the above equation we can solve for the optimal x1 as function of y (in general
it is also a function of w1 and w2 but remember we plugged numbers for these already):
1 1 2
p 3=2
3
= or x 1 = 2y ; or x^1(y) = 2y
x 21 2y3

which is the conditional input demand for input 1.

4. Now how do we nd the optimal quantity demanded of input 2 (i.e. its conditional input
3
demand, x^2)? Remember we know that x^2 = y x^1 so we have:

y3 y3=2
x^2(y) = p = p
2y3=2 2

which is the conditional input demand for input 2. Notice that both conditional input
demands depend on the given level of output, y that the rm wants to produce.

5. The minimum cost of producing y units of output then can be found by simply
substituting the conditional input demands obtained above into the cost function:
p 2 3=2 p 3=2
c^(y) = 1x^1(y) + 2x^2(y) = 2y3=2 + p y = 2 2y
2

Method 2 (using the optimality condition)

This method works even if it is not obvious how to express x2 in terms of x1 from the
constraint. However, it relies on tangency and interiority of the solution.

1. To be able to write down the optimality condition for cost minimization we need to
compute the marginal products of each input, i.e. the partial derivatives of the production
function with respect to x1 and x2: We have:
@f(x1; x2) 1 2=3 1=3
MP1(x1; x2) = = x 1 x2
@x 3
1

@f(x1; x2) 1 1=3 2=3

MP2(x1; x2) = = x 1 x2
@x 3
2

6
(remember when taking partial derivatives you hold all variables that you are not taking the
derivative with respect to as constant).

2. Thus the optimality condition for our particular production function and input prices is
(cancelling the 1/3s):
x 2=3 1=3
1 x2 1 x2 1
= or collecting the exponents, =
x1=3 2=3 2 x1 2
1 x 2

7
(please remember how to manipulate fractions of numbers raised to di erent powers!). From
above we get: x2 = 1 x2 1:

3. We can use the above relationship between x2 and x1 and plug it in the second equation that
they must satisfy, namely that output should be y :
1 1=3
x1=3
1 ( x1) =y
2
Again, raising both sides to power 3 helps:
1 2 p
x 1 = y 3 or, x21 = 2y3 ; so x^1(y) = 2y3=2
2

or the conditional input demand for input one is the same thing as before once again. Then
1 y3=2
x^2(y) = x^1(y) = p
2 2

again same as before. Clearly then the minimized costs will be also the same.

Method 3 (using Lagrange, OPTIONAL)

1. We write the Lagrangean for our speci c problem:
1=3 1=3
(x1; x2; ) = 1x1 + 2x2 (x1 x2 y)

2. Taking the partial derivatives with respect to the arguments of the Lagrangean we get:
1 2=3 1=3
1 x 1 x2 = 0
3
1 1=3 2=3
2 x 1 x2 = 0
3
x1=3 x1=3 y = 0
1 2

3. Re-arrange the rst two equations as described in the general example above and divide them
through to get:
2=3 1=3
1 1 x x2
= 1=3 2=3
2 x x
1 2
which is the same what we had in Method 2. Thus you can use the rest of the steps in
Method 2 to nish the solution.