The nature of inferior inputs

Akihito Asano∗

April 20, 2009

Abstract

We offer a simple model which helps to intuitively explain the firm’s decision making
when an inferior input is involved. The existence of inferior inputs is stated in Hicks’
(1946) seminal work, Value and Capital. A strand of research since has attempted
explaining a firm’s profit maximising behaviour when the inferior input is involved
in production. However, despite algebraic and diagrammatic proofs established in
the literature, economists still appear to be confused about the notion of the inferior
input, and the firm’s response to a change in the price of such an input is perceived
to be beyond intuition.

[Preliminary draft: Please do not quote]

Keywords: Regressive (inferior) inputs

JEL Classification: D24 (Production)

∗
School of Economics (Building 25A), The Australian National University, Canberra ACT 0200 AUS-
TRALIA. Tel: 02 6125 3363. Fax: 02 6125 5124. Email: [email protected]. The author would
like to thank Ben Smith for a fair amount of input as well as his criticism. The author also would like to
thank Shane Evans, Chris Jones and Rod Tyers for their constructive comments and useful discussion.

0
1 Introduction
A section of Hicks (1946), Value and Capital, has puzzled many economists.

Suppose the firm produces one product X, and employs two factors, A and B... the
demand for A must necessarily expand when its price falls... there are three ways in which
an expansion of the demand for A may be balanced:... (3) The demand for the factor B
may be reduced, but the supply of the product X may be reduced too... Let’s call it ‘re-
gression’. If the factor A and the product X are regressive, a substitution of A for B will
lower the marginal product of B in terms of X, and therefore (at given prices of B and X)
cause the supply of X to be contracted. I have a feeling that at this point the reader will
rub his eyes, and declare that something must have gone wrong with the argument...1

Such an input – called an inferior (or regressive) input – is, according to Hicks
(1946), anomalous, improbable and perverse. Many studies, however, have appreciated
its existence, and have attempted to uncover the nature of this mysterious type of input.
These include Ferguson and Saving (1969), Nagatani (1978), Puu (1971) Rowe (1977),
and Syrquin (1970). Takayama (1993) summarises the series of these studies and states
some strong results regarding the inferior input. Those include:

1. An increase in a factor price shifts the marginal cost curve down if the factor of the
focus is inferior.

2. An increase in a factor price increases the optimum output if the factor of the focus
is inferior.

3. The expansion (or scale) effect is always negative regardless of the factor of the focus
is normal or inferior, so the total effect is always negative.
1
Other two cases are: (1) The supply of the product X may increase, and the demand for the other
factor B may be reduced (here no complementarity is present). (2) The supply of X may be increased,
but the demand for B may increase as well (here the factors A and B are complementary).

1
 Yet, economists still appear to be confused about the notion of the inferior input.2 We
suppose the reason is because people still have trouble in believing in these results, which
are inter-related. That is, although some nice algebraic or diagrammatical explanations
have been provided, there still is no intuitive or convincing argument explaining why the
optimal output should rise in the event there is a rise in the price of an inferior input.
In the following, we attempt to intuitively explain a firm’s production decision when an
inferior input is involved.

2 The nature of the problem

We first review the algebraic proof by Nagatani (1978) to show the main statements con-
cerning inferior inputs are correct.3 Consider a price taking firm producing a commodity
whose market price is denoted by p. The firm employs n inputs x = (x1 , x2 , . . . , xn ) whose
price vector is denoted by w = (w1 , w2 , . . . , wn ).
We start with one of the duality properties in producer theory:

xi (p, w) = xhi (w, y(p, w)), (1)

where xi (p, w) and xhi (w, y(p, w)) are the input demand and conditional input demand
functions, respectively, for input i, and y(p, w) is the output supply function. Partially
differentiating Equation (1) with respect to wi we get,

∂xi (p, w) ∂xhi (w, y) ∂xh (w, y) ∂y(p, w)

= + i · (2)
∂wi ∂wi ∂y ∂wi
| {z } | {z }
substitution effect expansion effect

∂xi (p,w)
∂wi
shows the effect of the change in the price of input i on the demand for that
input. The equation say it can be decomposed into two effects shown on the right hand
side. The first term of the right hand side is the substitution effect. The second term
2
For example, Katz and Rosen (1998), in their diagram on p.319 entitled ‘The Output Effect May Be
Positive,’ demonstrate something completely orthogonal to the third result above.
3
Takayama (1993) has an extensive survey on this topic.

2
represents the expansion (or scale) effect. The substitution effect is always non-positive
due to concavity of the cost function. On the other hand, the expansion effect appears
∂xh
i (w,y)
rather ambiguous. The first component ∂y
depends on whether input i is normal or
inferior (regressive), i.e.
⎧
⎪ h
⎨ ∂xi (w,y) < 0 ⇐⇒ xi is inferior,
∂y
(3)
⎪
⎩ ∂xhi (w,y) > 0 ⇐⇒ xi is normal.
∂y

∂y(p,w)
If we know the sign of the second component ∂wi
we are done. To examine the sign of
it, we consider the profit maximisation problem:

max [py − C(w, y)] . (4)

The first order condition for the profit maximisation problem is:

∂C(w, y)
p = C 0 (w, y) ≡ , (5)
∂y

which merely states the price equals the marginal cost. Hence when wi rises, the optimal
y changes depending on how the marginal cost curve shifts, i.e.:

∂y ∂C 0 (w, y)
≶ 0 ⇐⇒ ≷0 (6)
∂wi ∂wi

In the meantime, invoking the Young’s Theorem and Shephard’s Lemma, we obtain

µ ¶ µ ¶
∂C 0 (w, y) ∂ ∂C(w, y) ∂ ∂C(w, y) ∂xhi
= = = . (7)
∂wi ∂wi ∂y ∂y ∂wi ∂y

Combining this equation with the inequality (6) we get,

µ ¶ µ ¶
∂y ∂xhi
sign = −sign . (8)
∂wi ∂y

3
Using this and the decomposition in Equation (2) we get the result we want:

µ ¶ µ ¶2
∂xhi (w, y) ∂y(p, w) ∂xhi
sign = −sign < 0. (9)
∂y ∂wi ∂y

Equation (8) shows that if an input of the focus is inferior, a rise in the price of the
input leads to an increase in the optimal output. This leads to the result in Equation (9),
which says that the sign of the expansion effect is, regardless of the type of input, always
negative (as long as it is not neutral, in which case the effect is zero), because the sign
of two components are shown to be always opposite. The three statements provided in
Introduction hence have been proven.

x2
6

d
c
T
T
b T
T r
T D

r
a Z T
r
Z CT B
Z T
Z T
Z
T
Z
Z
T
r
y0 + dy
TZZA
T Z
TT ZZ y0 -
O x1

Figure 1: An input is inferior when the marginal cost falls after a rise in the price of that
input

We describe these events diagrammatically for n = 2 in Figure 1. Start with the

situation where the firm is profit maximising (and so is cost minimising in producing y0 )
at Point A. Two isoquant curves corresponding to y0 and y0 + dy imply the marginal
cost, measured in input 2, is the distance ab. Now if the price of input 1 rises, holding

4
other prices constant, the cost minimising input combination in producing y0 changes
from Point A to Point C. Under the new input price ratio, the marginal cost is now
measured by the distance cd. If the isoquant curves shape as in Figure 1, it is possible
for the marginal cost to fall following an increase in the price of an input.
Note that in the figure, x1 is an inferior input so the (conditional) demand for it de-
creases as the level of output. Namely, under the original (new) input price ratio, the cost
minimising input combination changes from Point A (C, respectively) to Point B (D, re-
spectively) for the marginal increase in output, where less x1 is employed. Nagatani (1978)
demonstrated using some principles of geometry cd < ab if and only if Point A (C) is to
the right of Point B (D). It is his contribution that the phenomenon is explained without
the use of calculus, which helps many instructors who teach undergraduate students.

3 Let’s think about ‘regression’: Why does the MC

fall after an increase in the price of an input?
Any economists – even those who have not been exposed to the notion of inferior inputs –
by now agree that the established results concerning inferior inputs are correct. However,
how do they go about explaining the phenomenon? Why should the firm wait for an
increase in the price of an inferior input in order to increase its output level? Many would
say, ‘Why didn’t this firm increase the output in the first place?’ Both the algebraic
proof and diagrammatic explanation demonstrate the way a firm should respond to a
change in the price of an inferior input, but they do not intuitively explain its response.
It still remains as one of the results in undergraduate microeconomics material that is
hard to believe. One cannot find a persuading intuitive explanation even in prominent
intermediate microeconomics textbooks such as Katz and Rosen (1998) and Nicholson
(2007) or in a more advanced microeconomics textbook such as Silberberg and Suen
(2001).
The puzzle seems to stem from the fact that the optimal output level increases despite
the fact one of the input prices increases. Why didn’t the firm produce more before an

5
increase in the input price? Since the rise in the optimal output is caused by a fall in the
marginal cost, it may be a good idea to ask the following question, ‘What is necessary for
the marginal cost to decrease when the price of one of the inputs rises?’ As we can see
from Figure 1, it has to do with the firm’s cost minimisation problem, which is our next
focus.
For simplicity, suppose there are only two inputs x1 and x2 , where f (x1 , x2 ) is a
firm’s production function.4 We assume the marginal product for each input is positive
and diminishing. We consider the firm’s cost minimisation problem where λ denotes the
Lagrange multiplier, which is identified as the marginal cost. We would particularly like
to know the comparative static result when there is a marginal change in one of the
input prices. Since the results concerning the cost minimisation problem are standard
(unlike those related to inferior inputs) we shall just list them without proofs.5 If we let
∂2f
fij = ∂xj ∂xi
and
¯ ¯
¯ ¯
¯ −λf11 −λf12 −f1 ¯
¯ ¯
¯ ¯
H = ¯ −λf21 −λf22 −f2 ¯ , (10)
¯ ¯
¯ ¯
¯ −f1 −f2 0 ¯

we obtain the following:

∂λ H13
=− , (11)
∂w1 H

where H13 is the signed cofactor of the element in row i and column j of the determinant
H. The sufficient second order condition for the cost minimisation requires H < 0, so for
the marginal cost to fall when w1 increases, the sign of H13 needs to be negative. The
sign of H13 is:
4
The following argument should be able to be extended to the general case where there are more than
2 inputs.
5
For the details, see for example Silberberg and Suen (2001), Chapter 8.

6
 H13 = λ(f21 f2 − f22 f1 ) Q 0. (12)

As can be seen above, because the marginal product of each input is positive and dimin-
ishing, the only way the sign of H13 to be negative is to have f21 < 0. Now we have
reached the results established by Puu (1971) and Rowe (1977): a necessary, but not
sufficient, condition for an increase in the input price to lead to an increase in output is
that f12 = f21 is negative. Hicks (1946) call these inputs substitutes in demand.6 It is
straightforward to see why it is not a sufficient condition: in Equation (12) if the absolute
value of f22 f1 is not less than that of f21 f2 , then H13 is non-negative. For it to be negative,
these inputs need to be substitutes in demand to a fair extent.
What links between the fact the inputs are substitutes in demand (or rivals) and one
of them may be inferior? Also why is it more likely that one of them becomes inferior
when the two inputs are substitutes in demand to greater extent? We suggest a model to
further investigate these questions, which helps to intuitively explain the firm’s response
to a rise in the price of an inferior input.

4 An illustrative model
Suppose a firm (university) is producing a composite product Z, whose value is denoted
by V . The composite product comprises research R and degrees D, i.e.

Z = Z(R, D) (13)

and
V = pR R + pD D, (14)
6
This necessary condition is highlighted in Epstein and Spiegel (2000) where these inputs are called
rivals. They come up with a production function that allows for an inferior input, and shows that their
function is consistent with inputs being rivals. In commenting on their work, Weber (2001) correctly
points out that if a utility function allow for inferior consumer’s good, a production function with the
same functional form allows for an inferior input.

7
where pR and pD are the prices of research and degrees, respectively.
We assume there are two inputs, associate lecturers A and casual tutors C. Associate
lecturers can contribute to producing both research and degrees. If they are hired as
an associate lecturer, they spend half of their working time conducting research and the
other half delivering lectures.7 Since they can teach as well, they can be hired as a casual
tutor if they want. Casual tutors, in contrast, do not contribute to research and spend
all their time on teaching.
The trick to get intuition of an inferior input is to express everything in terms of time.
Firstly, denoting the times spent on research and teaching (degree production) activities
by tR and tD , respectively, we can express the above production functions in terms of time
spent on each of the activities: R = R(tR ) and D = D(tD ), where marginal products are
diminishing, i.e. R00 (tR ) < 0 and D00 (tD ) < 0. Secondly, suppose the exogenously given
t t
prices of “research time” and “teaching time” are wR and wD , respectively. Since casual
t
tutors only teach, the price of a casual tutor (per time) wC is equal to wD .

t
wD = wC . (15)

As for associate lecturers, their work is split into teaching and research equally. From
teaching, they get paid 12 wD
t
. From their research they get paid 12 wR
t
, and so the price of
1 t t
an associate lecturer (per time) wA is equal to 2
(wD + wR ). Therefore:

t
wR = 2wA − wC . (16)

Now, we can formally state a price-taking firm’s problem as the following:

£ t t
¤
max pR R(tR ) + pD D(tD ) − wR tR − wD tD . (17)
tR ,tD

The first-order necessary conditions (assuming both tR > 0 and tD > 0) are:
7
This is not an unrealistic assumption. Later we examine the effect of splitting an associate lecturer’s
time differently.

8
 pR R0 (tR ) = wR
t
, (18)

and
pD D0 (tD ) = wD
t
. (19)

These equations merely show that the value of marginal product of time spent of each of
the activities has to equal the price of the corresponding activity. In the following we first
investigate the effect of output expansion, which illustrates the nature of inferior inputs.
Our investigation is then related to the link between change in the price of an inferior
input and the marginal cost discussed earlier.

4.1 Output expansion

Suppose equi-proportional increases in the prices of research and degrees, so that the value
of marginal product (V MP ) schedules for both research and teaching shift upwards. The
firm now produces more composite output as it devotes more times for both research and
teaching. However, one of the inputs may be employed less. The situation is depicted on
two diagrams in Figure 2.
t t
If wR À wD , and recalling that the shifts are not vertically parallel, we can envisage
a case where the optimal “research time” increases by a significant amount but that
of “teaching time” increases by only a small amount. In the figure, the research time
increases from t0R to t1R , but the teaching time increases only from t0D to t1D . The only way
to increase the former is to hire more associate lecturers (from t0R to t1R ), but since they
teach half-time as well, this in turn may imply sacking some casual tutors. On the right
diagram in Figure 2, associate lecturers’ teaching time has increased from t0DA to t1DA ,
but casual tutors’ teaching time has decreased from t0DC to t1DC , despite an increase in
the aggregate teaching time. This is the case where the casual tutors are inferior inputs,
i.e. when the production increases holding the input prices constant, the demand for the

9
input decreases.

t t
wR wD
6 6

Z
Z
t Z r r
wR Z
Z
Z HH
Z HH
Z HH
Z
Z H
HH
r r
Z
Z t HH
Z V MPR1 wD HH V MPD1
Z
Z H
V MP 0 H V MP 0
-R -D
O t0 - t1 tR O t0D t1D tD
- -
R R
¾ t0 A -¾ t0DC
D
¾ t1 A -¾ t1 C -
D D

Figure 2: Inferior input can undertake only one task

4.2 Inputs that are substitutes in demand

We can relate the above events to the necessary condition – inputs must be substitutes in
demand for one of them to be inferior – we mentioned in the previous section. For casual
tutors, who can only teach, the more associate lecturers the university hires, the lower
their marginal product will become at the margin because these associate lecturers can
teach as well. For the associate lecturers, their marginal product of teaching goes down at
the margin as well with an increase in casual tutors for the same reason. Although their
research productivity is unaffected by the presence of casual tutors, their overall marginal
product will be negatively affected. Therefore the two inputs, associate lecturers and
casual tutors are substitutes in demand in our model.
As emphasised before, however, this is only a necessary condition for an input (in this
case, casual tutors) to be inferior. For an input to be inferior, we have mentioned it need
to be substitutes in demand to the other input to a fair extent. To get intuition of what

10
it means, we consider situations in our model under which a casual tutor is more likely
to be inferior.
It appears a sufficiently large gap between the productivity levels between research
and teaching – which is reflected in the gap between the prices paid to the two activities
– makes it more likely that a casual tutor is inferior. The productivity gap ensures a
sufficiently large increase in the research activity (distance t0R t1R in Figure 2) relative to
that in the teaching activity (distance t0D t1D in Figure 2) following the equi-proportional
price increases.

Remark 1. The greater the gap between the prices paid to the two activities, the more
likely casual tutors are inferior.

We can find a sufficient enough gap between the prices of research time and teaching
time for casual tutors to be inferior, given the way associate lecturers split their time
between two activities. It is a crucial assumption we imposed in our model that associate
lecturers are forced to split their time equally to two activities, and the way they split
their time appears to have an implication on whether casual tutors are inferior or not.
For example, suppose associate lecturers are forced to devote two-thirds of their time
to research and a third to teaching. Figure 3 is depicted exactly as Figure 2 except for
the difference in the split rule of the two activities. To the same equi-proportional price
increases discussed above, the firm increases both activities as in Figure 2 in terms of
time. However, the teaching activity is now shared differently by the two types of inputs.
Following the change in the output prices, the firm increases tR , which automatically
increases the time associate lecturers spend on teaching from t0DA to t0DA . The increase
is smaller if the proportion of the time they spend on teaching is smaller. The right
diagram in Figure 3 is drawn so that an increase in tDA is less than an increase in tD ,
the aggregate time spent on teaching. It is the case where the firm employ more of both
associate lecturers and casual tutors, in which case casual tutors are not inferior.8
8
It is implicitly assumed here that the V M P schedules do not shift following the change in the
proportion of times spent on the two activities. This issue is discussed in the next section.

11
 t t
wR wD
6 6

Z
Z
t Z r r
wR Z
Z
Z HH
Z HH
Z HH
Z
Z HH
r r
Z HH
Z w t
H
Z V MPR1 D HH V MPD1
Z
Z H
V MP 0 H V MP 0
-R ¾-¾ - -D
O t0R - t1R tR O t0 0 t 0 1
t tD
DA tD C -D
D

t1
¾DA
-¾
t1DC -

Figure 3: An input may not be inferior even when the other input is substitutes in demand

Remark 2. The greater the proportion of time associate lecturers spend on teaching, the
more likely casual tutors are inferior.

Remark 2 makes sense as the teaching activity is taken by both associate lecturers
and casual tutors. If more teaching is taken by associate lectures, that will force away
more casual tutors, making them an inferior input. Remark 1 essentially makes the same
point: if the wage gap between the two activities is greater, an equi-proportional price
increases in the prices of research and teaching enhance the research activity more than the
teaching activity, and so associate lecturers tend to encroach more on what casual tutors
specialise, making them an inferior input. Whilst encroachment by associate lecturers on
the teaching activity occurs in any case, these two remarks illustrate the encroachment
has to be sufficiently large to make casual tutors inferior. This is the gist of the necessary
condition established in the literature.

4.3 Marginal cost?

Using our model, we now investigate the biggest puzzle: the marginal cost falls when
a price of an inferior input rises hence the firm increases output. Suppose an increase

12
in the price of a casual tutor, wC , holding other things including wA constant. From
t
Equation (15) we know wD increases by the same amount. If associate lecturers split
t
their times equally to the two activities, Equation (16) suggest it be the case wR has
fallen by the equivalent amount. It makes sense because associate lecturers’ wage reflects
what they are paid for being engaged in the two activities.
In Figure 4 the effect of an increase in wC is depicted. As explained above the price of
the teaching activity rises but the price of the research activity falls, so the time the firm
chooses to spend on the former falls but the time it spends the latter rises. The value
of the composite good has risen by area A but has fallen by area B. Therefore, despite
an increase in the price of an input, we can think of a case where the total value of the
composite good produced increases. Since the output in our model is a composite good,
we cannot compare the output level itself before and after the input price change as in
the literature; this is an important point that we will come back and discuss.

t t
wR wD
6 6

Z
Z
t,0 Z r
wR Z
Z
t,1 ? Z r HH
wR Z HH
Z
Z t,1 HHr
Z wD HH
r
A Z 6 HH
Z t,0
wD H
Z HH
ZZ B
H
V MP H V MP
-R -D
O t0R - t1R tR O t 1¾
D t 0
D tD

Figure 4: The effect of a rise in the price of an inferior input: the output value rises

Suppose it is the case that area A is greater than area B in Figure 4 so the value of
the total output increases. To see what might have occurred to the marginal cost, we
consider the cost minimisation problem of the firm. Namely we ask how the firm would
t t
respond to an increase in wD (and a decrease in wR ) in Figure 4, if it were to produce

13
the same value of the composite good. Note we define the cost minimisation problem
differently to its traditional form because of the composite nature of our output.
The firm will increase the research activity and decrease the teaching activity to reduce
the cost. Consider a marginal increase in the research activity, which leads to an increase
in the value of the composite good. So we consider a marginal decrease in the teaching
activity to exactly offset the increase in the value of the composite good in order to hold
it constant. The firm continues the process so long as it can save the cost, so it must be
the case that the cost minimising combination of the two activities is (tcR , tcD ) in Figure 5.
The two bold rectangles (including the black rectangles) are the marginal values of the
composite good at the margin of the cost-minimising point, and hence their areas are
identical; the two bold rectangles’ areas are also equivalent.

t t
wR wD
6 6

Z
Z
t,0 Z r
wR Z
Z
t,1 ? Z r HH
wR Z HH
Z
Z t,1 HHr
Z wD HH
r
Z 6 HH
Z t,0
wD H
Z HH
ZZ H
V MP H V MP
-R -D
O t0R tcR t1R tR O tcDt1D t0D tD

Figure 5: The marginal cost of value production decreases if the price of an inferior input
rises

The readers must have already realised that we also need to consider the marginal
cost differently because of the nature of our output. The marginal cost of producing a
dollar worth of the composite good – not a unit of the good – before the input price
change is (trivially) a dollar, because the firm is profit maximising: the marginal cost of
value production (MCV P ) must be equal to the value of the marginal product at the

14
optimum.
After the price change the firm adjusts the combination of the two activities to cost-
minimise, holding the value output constant. The marginal cost of producing a dollar
worth of the composite good is the time cost of the research and teaching activities
required. As can be seen in Figure 5, after the adjustment made by the firm to cost-
minimise, both the marginal products are above their corresponding (new) prices. It
means the marginal cost of value production is less than unity, i.e. the marginal cost
decreased following an increase in an input price. The firm further adjust its input
combination: it will increase both research and teaching times to increase the total value
of the composite good.9
The marginal cost of value production falls, despite an increase in the price of the
teaching activity, because the price of the research activity falls. Recall the firm produces
a composite good – consisting of research and degrees – which has forced us to define
various notions differently from those in the literature. The firm can change the mix of
research and degrees following the input price change – in this case it starts producing
a more research oriented good – which implies the good produced after the input price
change is not exactly the same as what they produced before. In the case we have just
discussed above, the firm changes the mix of research and degrees after the input price
change, whilst holding the total value of the composite good constant. It then finds
producing more of the composite good, with even more research orientation, profitable.
It is likely to occur when research is much more productive than teaching and/or when a
fall in the price of research activity is large enough.
We can visualise that the fall in the price of the research activity must be sufficiently
large for the value of the marginal cost to fall. Figure 6 depicts the same situation as in
Figure 4 except now it is assumed associate lecturers spend their time twice as much on
t
research as they do on teaching. That is, we have wD = wC and:

t 3 1
wR = wA − wC . (20)
2 2
9
Note that an increase in associate lecturers crowd out some casual tutors during the process.

15
In this case, a dollar increase in the price of the teaching activity lowers the price of the
research activity by only a half dollar, as associate lecturers spend only a third of their
time in the teaching activity. As a result, the teaching activity increases from t0R to t2R .
As one can see, whilst the total value of the composite good decreases by area B as in the
previous case, it only increases by area C, which is strictly smaller than area A. As we
assumed in the previous subsection, if casual tutors ceased to be inferior once associate
lecturers spend two thirds of their time in research, it means area C is smaller than area
B. One can easily describe the situation where the productivity difference between the
two activities is not sufficiently large to allow for inferiority of casual tutors (which is
consistent with Remark 1).

t t
wR wD
6 6

Z
Z
Z r
Zr
t,0
w Z
t,2 R ?
wR Z HH
Z HH
Z
Z t,1 HHr
Z wD HH
r
C Z 6 HH
Z t,0
wD H
Z HH
ZZ B
H
V MP H V MP
-R -D
O t0R-t2R t1R tR O t1 ¾D t0 D tD

Figure 6: The effect of a rise in the price of a normal input: the output value falls

In this case, following the input price change, the firm adjusts the composition of
research and degrees, whilst holding the value of the composite good constant. It involves
increasing the time spent on research and decreasing the time spent of teaching. In
Figure 7, the cost-minimising combination of the two activities is shown. As in Figure 5,
the two bold rectangles (including the black rectangles) are the marginal values of the
composite good at the margin of the cost-minimising point, and hence their areas are

16
identical; the two bold rectangles’ areas are also equivalent. At the cost minimising
margin, both the marginal products will be below their respective prices, i.e. the marginal
cost of value production is above unity.
The firm, as in the previous case, does produce a more research oriented composite
good after the input price change. However the fall in the price of the research activity is
not large enough to justify that it produces a greater value of the composite good.

t t
wR wD
6 6

Z
Z
Z r
Zr
t,0
w Z
t,2 R ?
wR Z HH
Z HH
Z
Z t,1 HHr
Z wD HH
r
Z 6 HH
Z t,0
wD H
Z HH
ZZ H
V MP H V MP
-R -D
O t0R t2R tcR tR O t1DtcD t0D tD

Figure 7: The marginal cost of value production rises if the price of a normal input rises

5 Discussion
The key to our argument is one of the two inputs can be used for two activities – for
that reason, hereafter let us call this input a superior input – and the other one is used
only for one of the two activities, in which case, it may be inferior. Bearing this in mind,
we review the argument regarding an inferior input provided by Katz and Rosen (1998):

Think about a manufacturer that is planning to scrap its current factory and build a
new factory in order to increase its production level. If the increase in output is substan-
tial, then the firm may choose to build a highly automated factory requiring almost no

17
workers. Thus, even though output has increased, the quantity of labor hired will go down.

The Katz and Rosen (1998) example is a realistic description of a labour input, and
indeed, in view of our analysis in the previous section, it does make sense. The point is
to think that capital (or machines) may be superior in the sense that not only can they do
what labour does not do, but they can also do what labour does.
Suppose an increase in production, which is driven by an increase in the price of the
output. In increasing the production level, it may turn out more profitable for the firm
to increase capital but to decrease capital, because by employing new machines, which
can undertake the task labour undertakes, might make some of the existing labour force
redundant. This is the case where labour is an inferior input.
One crucial assumption we have made concerns the proportion of the time a superior
input allocates to the two activities – it is fixed – and it has an implication on whether
or not the other good becomes inferior. In reality, this proportion is also firms’ choice
variable, but perhaps to a lesser extent in the short-run. The proportion may be fixed
due to contracts in our example. In the capital and labour example à la Katz and Rosen
(1998), changing the proportion implies replacing the old machines with the new ones,
which may not be considered a short-run decision. In other words, the Katz and Rosen
(1998) argument is valid so long as the newly hired machines are of the similar type to
the existing ones. Then, the part of their argument, the firm may choose to build a highly
automated factory, becomes a little problematic.
To discuss this matter further, let us assume that capital and labour are the two inputs,
where capital is superior in the sense it can undertake two tasks – the heavy task and the
light task – which are required to produce the output. On the other hand, labour can
only conduct the light task. Suppose the firm currently makes the use of machines that
undertake equal (time) amount of heavy and light tasks when they are in operation. This
corresponds to our university example, in which associate lecturers are forced to allocate
equal times to teaching and research.
Suppose that, due to innovation or something, a new type of machine has become

18
available. To be consistent with the Katz and Rosen (1998) example, we postulate that
the new type of machine, when it is in operation, undertakes twice as much the light task
as the heavy task. That is, if the firm wishes to do one hour of the heavy task, two hours
of light tasks are undertaken by the machine as well. This leads to a highly automated
factory. Presumably, the new machine is more productive in the heavy task leading to an
upward shift in the value of marginal product schedule for the heavy task, V MPH , which
is described in Figure 8.
t
wH
6

Z
Z
Z
t Z r r
wH Z
Z
Z
Z
Z
Z
Z
Z
Z
Z V MP 1
Z
Z
H
0
V MP
-H
O t0H - t1H tH

Figure 8: Shifting the V MPH schedule

The shaded area in Figure 8 represents an increase in the profits for this firm.10 Since
in the market for the time for the light task, nothing but the composition of the two
inputs has changed – machines are spending more time on the light task and labour is
spending less time as a consequence – as long as the cost of upgrading the machines does
not exceed this increase in the profits, the firm will implement the upgrade. Therefore,
the optimal technology for the firm will be determined by equating the marginal benefit
and the marginal cost of upgrading, assuming that the former diminishes and the latter
10
We are assuming that labour is not completely replaced by these new machines.

19
increases.11 If the upgrade occurs, some of the labour becomes redundant and at the same
time the output increases.
By upgrading the machines – in the Katz and Rosen (1998) wording, to build a highly
automated factory – the readers might envisage the firm is now producing something
slightly different from what they used to produce. In the existing theoretical literature
where the product has a unique nature – the production function is merely f (x) – it is
difficult to imagine that producing more output is optimal when there is an increase in
the price of an input. Both the example by Katz and Rosen (1998) and our model focus
on composite products whose nature might differ depending on how they are produced.
Considering the composite nature of the product as well as thinking inputs in a different
dimension – e.g. tasks, times etc. – help us understand the behaviour of a firm that
utilises an inferior input.

References
[1] Epstein, G.S. and Spiegel, U. (2000) “A Production Function with an Inferior Input”
The Manchester School 68, 503—515.

[2] Ferguson, C.E. and Saving, T.R. (1969) “Long-Run Scale Adjustments of a Perfectly
Competitive Firm and Industry” American Economic Review 59, 774—783.

[3] Hicks, J.R. (1946) Value and Capital, 2nd Edition, Claredon Press: Oxford.

[4] Katz, M.L. and Rosen, H.S. (1998) Microeconomics, 3rd Edition, Irwin/McGraw-Hill:
Boston.

[5] Nagatani, K. (1978) “Substitution and Scale Effects in Factor Demands” Canadian
Journal of Economics 11, 521—527.
11
The former may be increasing, but even in that case, the state-of-art technology sets the upper bound
of the V M PH schedule.

20
[6] Nicholson, W. and Snyder, C. (2007) Microeconomic Theory: Basic Principles and
Extensions, 10th Edition, Thomson: New York.

[7] Puu, T. (1971) “Some Comments on “Inferior” (Regressive) Inputs” Swedish Journal
of Economics 73, 241—251.

[8] Rowe, J.W. (1977) “Some Further Comments on Input Classifications” Scandinavian
Journal of Economics 44, 488—496.

[9] Silberberg, E. and Suen, W. (2001) The Structure of Economics: A Mathematical

Analysis, 3rd Edition, McGraw-Hill: New York.

[10] Syrquin, M. (1970) “A Note of Inferior Inputs” Review of Economic Studies 37,
591—598.

[11] Takayama, A. (1993) Analytical Methods in Economics, The University of Michigan

Press: Ann Arbor.

[12] Weber, C.E. (2001) “A Production Function with an Inferior Input: Comment” The
Manchester School 69, 616—622.

21