John Riley 26 July 2019

Single variable optimization

This first module reviews the basics of one variable maximization from an economics
perspective. Consider a monopoly which produces a single product. The cost of producing each
unit is c. Current output is q0 . You have been recruited to figure out how much the firm should
produce and hence what price to set. What should you do? As a well trained economist you
seek information from the marketing department on how the demand for the product would
vary with price. You then use a statistical software package1 to estimate a demand price
function, p(q) . In the example depicted below the estimated demand price function is linear.
You can now compute a revenue function, R(q)  p(q)q .

Fig.1: Estimated demand price function

You pick some new output q1  q0 q where q is small and then compute the rate at which
revenue and cost rises with output.

R(q1)  R(q0 ) R(q0 q)  R(q0 )

 .
q1  q0 q
C(q1)  C(q0 ) cq1  cq0
 1 0 c
q1  q0 q q

Profit is (q)  R(q) C(q) . Thus the rate at which profit rises with output is

1
For this purpose Excel’s statistical package would be more than adequate.

1
John Riley 26 July 2019

(q1) (q0 ) (R(q1)  C(q1))  (R(q0 )  C(q0 )) R(q1)  R(q0 ) C(q1)  C(q0 )
  
q1  q0 q1  q0 q1  q0 q1  q0
Therefore

(q1) (q0 ) R(q1)  R(q0 )

 c
q1  q0 q1  q0
If current output is q0 and the rate at which revenue rises with output is strictly greater
than c, then profit can be increased by increasing output. Thus as a consultant you can advice
this course of action. Conversely, if the rate at which revenue rises with output is strictly less
than c, then profit can be increased by decreasing output.
Of course the next question is how much should output be changed. As a practical
matter you compute the rate at which revenue rises with output for a range of outputs. In the
example depicted below p(q)  60  401 q , the unit cost is 20, q0  200 and q 100 .

Fig. 2: Rate of increase in revenue and cost

The rate at which revenue rises with output is higher than 20 for all q  700 and lower for all
q  800. Thus profit is maximized at an output somewhere between 700 and 800.

2
John Riley 26 July 2019

Since the rate at which revenue rises is close to 20 on this interval, as a practical matter
it is not necessary to reduce q to something very small in order to get a more accurate
answer. Remember that the demand price function is an estimate based on the judgment of
the marketing department.

Derivative of a function

However, for a deeper economic analysis it is very helpful indeed to have an exact
solution to the maximization problem. Consider the expression for the rate of change of the
firm’s cost once more.

R(q1)  R(q0 ) R(q0 q)  R(q0 )

 .
q1  q0 q
right hand side has a limit as q approaches zero.

As an example, suppose that the estimated demand price function is linear

p(q)  a bq .
Then the revenue function is R(q)  p(q)q  aq  bq2 . Then

R(q0 q)  a(q0 q)  b(q0 q)2  aq0  b(q0 )2  aq0q  2bq0q  b(q)2 .
Subtracting R(q0 ) and dividing by q , the rate at which revenue rises with q is

R(q1)  R(q0 ) R
  a  2bq0  bq .
q q
1 0
q
Note that the expression on the right hand side has a limit as q approaches zero. Then, as an
economist would say, the “marginal revenue” of the firm is

MR(q0 )  a  2bq0
R
To a mathematician, if the ratio  a  2bq0 has a limit at q0 , the revenue function is
q
“differentiable” at q0 and the limit itself is the derivative of the revenue function at q0 .

The derivative of a function is typically written in one of the following three ways.

3
John Riley 26 July 2019

dR
(q)  R(q)  D R(q)
dq
Marginal cost is defined in the same way.

C(q1)  C(q0 ) C(q0 q)  C(q0 ) cq

  c .
q1  q0 q q
Since the ratio is c for all q  0, the limit of this ratio is c.

The marginal profit of the firm at output level q is therefore

d dR dC a c
(q)  (q)  (q)  a  2bq  c  2b(  q)
dq dq dq 2b
Note that marginal profit is a decreasing function of output. It is depicted below.

a c a c
Marginal profit is strictly positive for any q  and strictly negative for any q  . Thus
2b 2b
a c
profit is maximized by choosing the output q*  . Since (a, b, c)  (60, 401 ,20) the profit
2b
maximizing output is 800.
There is a second helpful way of thinking about the problem geometrically. In the
diagram on the left below we have drawn the line through the points (q0 , R(q0 ))  (200,10000)

4
John Riley 26 July 2019

R(q1)  R(q0 )
and (q1, R(q1))  (1000,35000) . The slope of this line is . Thus the equation of
q1  q0
the line is

R(q1)  R(q0 )
R  R(q0 )  (q  q0 ) .
q q
1 0

In the diagram on the right the increase in output, q  q2  q0 is smaller. The equation of the
new line is

R(q2 )  R(q0 )
R  R(q0 )  (q  q0 ) .
q q
2 0

Taking the limit as q approaches zero yields the line

dR 0
R  R(q0 )  (q )(q  q0 ) .
dq

The slope of this line is the marginal revenue at q0

Graphically this line touches the revenue curve at q0 . Mathematicians call this a tangent line.

5
John Riley 26 July 2019

In the next figure the graph of the profit function is depicted and also the tangent line at
q0  200 and the tangent line at q*  800 . Since the derivative of the profit function (marginal
profit) is zero at q* the slope of the tangent line at q* is zero.

6
John Riley 26 July 2019

Strict local maxima and minima

It is helpful to define (a, b) to be the “open interval” of real numbers q satisfying
a  q  b . If we add the boundary points this becomes a “closed interval” written [a, b] .
df
Suppose that (q) is zero at q0 . Suppose also that marginal profit is strictly positive for all q
dq
in the interval (a, q0 ) and strictly negative in the interval (q0 , b) . This is depicted below.

Fig. 3: Local Maximum at

It follows that f (q) is strictly larger at q0 than at any other point in the interval (a, b) .

When this is the case for some interval (a, b) containing q0 we say that f (q) has a strict local

maximum at q0 .

A strict local minimum is similarly defined.

Suppose, for example that the profit function of a firm is

(q) 192q 88q2 16q3  q4

Marginal profit is therefore

(q) 192 176q  48q2  4q3  4(48  44q 12q2  q3)

7
John Riley 26 July 2019

As you may confirm2, marginal profit can be rewritten as follows:

d
 4(q  2)(q  4)(q  6)
dq
Note that marginal profit is zero if output is 2, 4 or 6. As you can confirm, marginal profit
changes sign at each of these outputs. For example, marginal profit is strictly negative in the
interval (2,4) and strictly positive in the interval (4,6) . Thus q0  4 is a local minimum. The
profit function is depicted below. It has two local maxima.

Fig. 4: Profit function with two local maxima

It is not clear from the figure which of these solves the maximization problem. The answer can
then be obtained by substituting the two local maxima into the profit function.

2
Hopefully I did the algebra correctly!