44

The consumer
(I)
general
preference relation also is perfectly known. Contrary to Samuelson's
suggestion, therefore, it is possible to determine this relation from direct
observation of consumer behaviour without having to confront the
consumer with a series of binary choices.
3
The producer
L Definitions
We come now to the activity of producers, also called 'firms'. This will be
investigated in two successive stages. First of all we shall study the representa-
tion of the technical constraints which limit the range of feasible productive
processes. We must then formalise the decisions of the firm which must act
within a certain institutional context. Our discussion will be carried on
mainly in the context of 'perfect competition', which cannot pretend to be
an always valid description of real situations. But it is the ideal model on
which the study of the problems of general equilibrium arising in market
economies has been based so far.
As in our discussion of consumptiQn theory, we shall omit the index j
relating to the particular agent considered. So a
h
, b
h
and Yh will simply
denote input, output and net production of the good h in the firm in question.
For the purposes of economic theory, a detailed description of technical
processes is as pointless as knowledge of consumers' motivations. All that
matters in this chapter is that we should formalise the constraints which
technology imposes on the producer. These can be summarised in a very
simple way: certain vectors Y correspond to technically possible transforma-
tions of inputs into outputs; other vectors correspond to transformations
which are not allowed by the technology at the disposal of the firm.
To take account of this, we need only define in Rl the production set Yas
that set containing the net production vectors which are feasible for the
producer. Thus the demands of technology are represented by the simple
constraint
Y E Y.
(We must not forget that Y relates to a particular producer; in
equilibrium theory, each producer j has his own set Y
j
.)
Of course, all the technically feasible transformations are not of interest
a pnon; some may require greater inputs and yield smalfer outputs than
others. The firm's technical experts must eliminate the former in favour of
the latter. This is why we can often confine ourselves a priori to technically
efficient net productions. By this we mean any transformation which cannot
be altered so as to yield larger net production of one good without this
resulting in smaller net production of some other good. Relative to such a
transformation, therefore, output of one good cannot be increased without
increasing input or reducing output of another good.
Formally, the vector yl is said to be technically efficient if it belongs to the
ll
set Y of feasible net productions and if there exists no other vector y2 of Y
such that
:? y/; for h = 1,2, ... , I.
So the technically efficient vectors y belong to a subset, or possibly to the
whole, of the boundary of Y in the commodity space.t
In the construction of optimum and equilibrium theories we could impose
on ourselves to use the production set Yas the sole representation of technical
constraints. This is the method adopted in the most modern approaches to
the subject. Following a tradition of almost a century, however, mathematical
economl'sts often introduce another more restrictive concept, that of the
'production function', which formalises in particular the idea that marginal
substitutions between inputs are feasible.
Actually, in their approach to the problems of general equilibrium econo-
mists have alternatively used two types of formalisations, which stress two
opposing features of production. One feature is the existence of 'propor-
tionalities' or 'coefficients of production': some inputs must be combined in
given proportions, like iron ore and coal in the process of producing pig
iron. Another feature is the possibility of substituting an input for another:
machines can replace men, one fuel can be substituted for another, more or
less fertilizer can be put in a given piece of agricultural land and more or less
labour can be spent on it, hence the same crop may be achieved with a little
less fertilizer and a little more labour.
Economists such as K. Marx or L. Walras in the first editions of his
treatise constructed their systems assuming fixed proportionalities, i.e. com-
plementarity between inputs. Others like V. Pareto have used formalisations
implying that substitutabilities are everywhere prevalent. The great advantage
of the modern set theoretic approach is to cover both complementarities
t Rigorously, we can confine ourselves to technically efficient vectors only if, corres-
ponding to every y of Y, we can find an efficient y* such that :> )'h for all h. This will be
the case if Y is a closed set and if, without leaving Y, we cannot increase one component of
y indefinitely without reducing another. It docs not restrict the validity of the theory to
assume this.
(3)
(2)
47
Dejinitiol1s
and substitutions. The definition of Y can take into accoun.t
the substitutability of machines for men and the proportIOnality
. and coal Hence the theory built directly on Y IS fully genera III t IS
Iron ore, .
respect . f
Whe;1 we want to build models that lend themselves to or
dealing with questions of applied economics, we have cholce
d

between two types of more specific either pro
functions, usually allowing fo.r large or fixed coeffiCIent
processes combined into 'actIVity analySIS models. d .
such as the present ones should not ignore th.e pro uctlOn
function concept. In fact it will be used extensively with aIm of
. . .' d to allow the free use of differential calculus. ome
cxposltlon easier an . Y b
essel1tJ
'al proofs will be given under the assumption that the se.ts .i can e
, . h h' tlOn IS not re-
represented by production functIOns, even th.oug t assump h' f b
'd for the validity of the result. ProductIOn functIOns must t ore. e
qUIre h 11 . t t 111 passll1g
deflned and discussed with some care. Later on we s a P01l1. ou
places where the use of such functions conceals some difficulty.
!:::..J?D2..duction {unction (for a particular finn is, by definition, a real function
defined on R
1
such that:
It]'!, Y2, ... , )'/) = 0
if and only if Y is an efficient vector, and such that
/(1'[,)'2, .··,Y/) 0
if and only if)' belongs to Y. . ". . fi d b
For the moment we shall not inquire Into the to be e .y
Y if weare to be able to define such a function. ThiS wIiI be dIscussed 111
Section 2. . . esent
According to this definition, we can use (I) or (3) eq.Ulvalentl
y
todrepr th
the technical constraints on productiont (the functIOn / depen s on e
particular producer j, as does Y). . .
Geometric illustrations of the production set and the productIOn
are often fruitful. Suppose, for example, that there are four commo lIes,
the first two of which are outputs of the firm and the last two Il1puts.
and 2 represent two intersections ,of Y, b
Y
=
, _ ;0) the second by a hyperplane (YI - J ! , Y2 . 2 .,
) 4 - ) 4 t' tl set of tJv' productions that are feasible from the quantllJes
rcpresen s 1e .
-t'- We may point out that, like tile Ulilit
y
] ffunccttiOonn"
. I F ,Ie If ¢ IS a rea un I
defined unique y. or examl, . then ¢(f) corresponds to the same set as f.
and which is zero when Its argumcndt IS tl . n consumption theory, we shall not lay
Since this has already been dlscusse sU 1Clen y I
further stress on it.
The producer 46
48
The producer Definit ions 49
The function g* will generally be increasing with the a
h
and thefunction g will
consequently be decreasing with respect to the Yh' or at least non-increasing.
Later on we shall often assume that the function f is twice differentiable.
Let yO and yO + dy be two neighbouring technically efficient vectors. We can
write
= - and a2 = - y2 of the two inputs; the second represents the set of
inputs allowing the quantities b? = Y? and bg = yg of the two outputs to be
obtained. The points satisfying (2) are represented by the North-Easi
boundary on Figure 1 and the South-West boundary on Figure 2. ryve note
'. in passing that a set which, like the curve in Figure 2, represents the technically
II
combinations of inputs yielding given quantities of outputs is called
an zsoquant.)
/
L dYh = 0
h=1
(8)
(9)
(11)
(12)
(10)
s =f. I
s, r =f. 1.
for
for
-
dYr g,;
- ==
dy', g;
and
The ratio (II) measures the increase in production resulting from an
increase of one unit in the input of s (note that Y
s
is equal to minus the input).
It is often called the marginal productivity of s. The ratio (12) defines, apart
from sign, the additional quantity of input of r which is necessary to compen-
sate in output for a reduction of one unit in the input of s. This is, in fact, a
marginal rate of substitution.
We note also that the first derivativesfh of the production functionfmust
take non-negative values at every technically efficient point yO. Consider a
small variation dy all of whose components are zero except dYk' which is
assumed positive. Since yO is technically efficient, yO + dy is not technically
possible, that is, f(yO + dy) is positive. But, since f(yO) is zero, f(yO + dy)
can be positive only if fk is not negative.
where.fh denotes the value at yO of the derivative of f with respect to Yh'
In particular, if all the dYh except two, dYr and dys' are zero, then (8) reduces
to
dYr
dy',
The ratio on the right hand side of (10) can be called theJnarginal rate of
substitution the producer in gueiliml, This
expression is similar to that encountered in consumption theory. To avoid
confusion, we shall sometimes speak instead of the !!!:.arginal r{lte of
formation.
In the particular case where f takes the form (4), equalities of the type
(10) become
or
Fig. 2

Fig. 1
*YI g(Y2' ... , Yf) (5)
and the expression 'production function' is also used for the function g
which defines the output resulting from given quantities of inputs. There
should be no real possibility of confusion from this ambiguity.
Note that we could show inputs and outputs explicitly in (5). Thus
bi g(- a2, - a
J
, ... , - aa (6)
or, after an obvious change in notation,
bl g*(a2, aJ, ... , a
z
)· (7)
The most general form of a production function is that in (2). Slightly
more particular expressions are often used. Thus it is often assumed that the
firm has only one output, the good I, to fix ideas; the production function
is then given the form:t
"·,Yf) = YI - g(Y2' ... ,y/). (4)
The technical constraint is
b • Y
2 2
t Obviously this particular form is no longer affected by the indeterminacy already
mentioned in relation to the general form (2), Here the function g representing a given
set Y is determined uniquely. In fact, even if these are several outputs, in most cases we
can solve the equality fCy) = 0 for YI and so revert to (5).
50 The producer
The validity of production functions 51
2. The validity of production functions
We must now investigate the conditions to be satisfied by the production
set Y in order that, first of all, there exists a production function/, and in the
second place, that this function is differentiable. These conditions are
certainly more restrictive than it would appear at first glance.
Differentiability implies that f is continuous and consequently that Y is a
, closed set in RI. This property is not restrictive; if the vectors {y\, y
2
, •.. ,}
of a convergent sequence each define a feasible production then the limiting
vector certainly corresponds in reality to a feasible production.
2
efficient vector of Yand thatf(y) > °is satisfied for every vector y outside Y.
(At a point such as N,/(y) should be equal to a num?er,
I
be positive point !y whose second coordll1ate IS pOSItIve; thIS
is incompatIble WIth the contll1Ulty off at N.)
However, we can take account of these limitations by altering the definition
of the production function and explicitly adding inequalities to the formal
representations of the set Y and the set of technically efficient vectors. For
example, to characterise Y we replace (3) by
{
[(Y\,YZ' ... ,y/) 0, (13)
Yh ° for a specified list of goods h.
To characterise the set of technically efficient vectors, (2) is replaced by
{
Yl + o:Y2 = ° (16)
yz 0.
This complication will not be taken into account in our discussion of the
general theories. That is, we shall proceed as if the limits on the domains J2f
variation of neveriil force. Aswe-sawli1-consumption theory,
new particular features are revealed if we take account of constraints
expressed by inequalities, but this does not alter basically the nature of the
results. We shall presently return to this point.
(ii) In the second place, in some productive operations the different goods
which constitute inputs must be combined in fixed proportions. This is
particularly the case for most of the raw materials used in many industrial
processes.
When such proportionality ratios exist, the isoquants do not have the
same form as in Figure 2. If there is free disposal of surplus, they look like
the isoquant in Figure 4. Apart from the surplus of one of the two inputs,
a
3
and a
4
must take values whose ratio corresponds to that defined by the
half-line GA. Except at the point A, the half-lines AN and AM correspond to
non-technically efficient productions. At the point A, the first derivatives off
with respect to Y3 and Y4 are not continuous. (The situation is similar to that
in Chapter 2, with the utility function (8) illustrated in Figure 7.)
The real situation is sometimes less clear-cut than Figure 4 aS5umes, since
M
Fig. 3
But the continuity off implies also that every point y* on the boundary of
Ysatisfiesf(y*) = 0 since it can be approached both by a sequence of vectors
y such thatf(y) 0 and by a sequence of vectors such thatf(y) > 0. So the 1\
definition off implies that every point y* on the boundary of Y is technically
efficient. Moreover, differentiability assumes that, with respect to any
technically efficient vector, the marginal rates of substitution are all well-
defined. Taken literally, these consequences are ditDcult to accept.
(i) In the first place, the domains of variation of all, or some, of the y" may
be limited. For example, technology may demand that some good r occurs
only as input and some other good s only as output. So the inequalities
Yr °and Ys 0 appear in the definition of Y. (In fact, the second in-
equality can be eliminated if we assume that the firm can always dispose of
its surplus without cost, since this assumption is naturally expressed as:
yO E Y and y" for all h implies y E Y.) Because of the limits on the
domains of variation of SOll,e y", the set Y has boundaries corresponding to
non-technically efficient )roductions (for example, the half-line GN in
Figure 3).
The existence of such boundaries is incompatible with the cOlltinuity of f
together with the conditions thatf(y) < °is satisfied for every non-technically
{
f(Yl' Y2, ... , y/) = 0,
y" ° for the same list of goods h.
Thus, for Figure 3, (13) and (14, become
{
y\ + o:yz 0,
yz 0,
and
(14)
(15)
52
The producer
Assumptions about production sets 53
t It is the aim of a new branch of economic science, 'activity analysis', to integrate into
the theory formalisations which describe technical constraints more accurately than do
production functions. A very good account of thc resulting modifications is given in
Dorfman, Application of Linear Programming to the Theory of the Firm, University of
California Press, Berkeley 1951. See also Dorfman, Samuelson and Solo'.", Linear program-
ming alld activity analysis, McGraw-Hili, New York, 1958.
3. Assumptions about production sets
We must now discuss certain assumptions which are frequently adopted
about production sets or production functions.
ADDITIVITY. If the two vectors y
1
and y
2
define feasible productions
(yl E Y and y2 E Y or f(y
1
) 0 andf(y2) 0), then the vector y = )'1 -+ y2
defines a feasible production (therefore y E Y or fey) 0).
This appears a natural assumption. For; it seems that we can always
realise y by realising independently yl and y2. Additivity fails to hold only
if yl and y2 cannot be applied simultaneously. A priori there seelTl$ no
reason for this to be the case.
However, it may happen that the model does not identify all the com-
modities which in fact occur as inputs in production operations. For example,
if the land in the possession of an agricultural undertaking does not appear
among the commodities, then additivity does not apply to its production set,
since, if the available land is totally used by yl on the one hand and by y2
on the other, realisation of yl -+ y2 requires double the actually available
quantity of land. Similarly, if the capacity for work of the head of an
industrial firm does not appear among the commodities, and if his capacity
limits production, then additi vity no longer strictly applies.
encountered in their rigorous exposition. The changes in production theory
introduced by their presence will be described briefly. t
Finally, we see that the above-mentioned difficulties can be avoided if we
base our reasoning directly on the set Y of feasible productions and on the
set of technically efficient productions rather than on the production function.
This is the approach adopted in the most modern treatments of the theories
with which we are concerned here.
As when a utility function is substituted for a preordering of consumer
choices, the substitution of a production function for a production set makes
exposition easier since it allows the use of the differential calculus and of
fairly standard types of mathematical reasoning. Moreover, this approach
alone leads to certain results which every economist must know. Knowledge
of these results is essential for the student, even if their application is some-
what restricted by the simplifications required to justify the production
function.
(17)
Fig. 5
o
. _
Fig. 4
fA M
8 _!/
3 3
Y4 = aY3'
In the case of two techniques, as in Figure 5, the supplementary constraints
may be
there may be available to the firm two or more production techniques each
requiring fixed proportions of inputs, the proportions differing for the
different techniques. Figure 5 relates to an example of two techniques, the
first represented by the point A, the second by the point B. The firm can
employ the two simultaneously to produce the same quantities of
outputs. For example, if each technique can be employed on a scale reduced
by one half relative to that represented by A or B (the assumption of constant
returns to scale, to be defined presently) then the same output can be obtained
by simultaneous use of the two techniques on this new scale; the point on
Figure 5 corresponding to this method of production is the midpoint of AB.
Similarly, each point on AB defines a possible combination of the two
techniques yielding the same output as A or B. In this case, the first deriva-
tives offare in fact continuous at each point within AB, but not at A nor at B.
In order formally to represent such situations as those of Figures 4 and 5,
we can add other constraints to the equation fey) = 0 to characterise the set
of technically efficient vectors. For example, if, as in Figure 4, there must be
a fixed proportion between Y3 and Y4' we write:
- f3YJ - Y4 - aYJ. (18)
The theory becomes very complicated if such constraints are taken into
account. For this reason, they are better ignored in a course of lectures
whose aim is to provide the student with a sound grasp of the general logic
of the theories to be discussed rather than the difficulties which are
54 The producer
Assumptions about production sets
55
DIVISIBILITY. If the vector yl defines a feasible production (yl E Y or
(fyl) :s;; 0) and if 0 < a < 1, then the vector ayl also defines a feasible
production (therefore ayl E Y and f(ayl) :s;; 0).
This assumption is much less generally satisfied than the previous one.
It assumes that every productive operation can be split up and realised on a
reduced scale without changing the proportions of inputs and outputs.
Taken literally, it can be said to be rarely satisfied. For every productive
operation there is certainly a level below which it cannot be carried out in
unaltered conditions. But this indivisibility may vary in its degree of effective-
ness and in many industrial operations it appears negligible.
CONSTANT RETURNS TO SCALE. t If the vector yl defines a feasible production
(yl E Y or f(yl) :s;; 0) and if f3 is a, positive number, then the vector f3y l also
defines a feasible production (therefore f3y l E Yandf(f3y l) :s;; 0).
,Obviously the constant defined by this assumption imply divisibility.
.#. and divisibility imply constant returns to scale.
let k be the integral part -of f3; we (;ill apply the-property oTadditivit;
repeatedly, taking the vectors yl, 2y l, ... , (k - I)yl successively for yZ and
thus proving that 2y\ 3y\ ... , ky l are feasible; divisibility shows that
(f3 - k)yl is feasible; finally, additivity shows that f3y l = (f3 - k)yl + ky l
is feasible.
In practice, we shall consider that returns to scale are constant precisely
when additivity and divisibility can be considered to hold, although
rigorously, additivity is not necessary.
Consider the particular case where the technical constraints are expressed
in the form (5). If the function 9 is homogeneous of the first degree, then the
assumption of constant returns to scale is clearly satisfied.
Conversely, constant returns to scale imply that
g(f3Yz, ... , f3y/) = f3g(yz, ... , y/)
for every vector Y and every positive number fJ. Indeed, on the one hand
the hypothesis implies, by definition,
g(f3yz, ... , f3y/) > f3Yl = f3g(yz, ... , y/),
since f3y is feasible whenever Y is feasible, On the other hand, the same
hypothesis implies:
g(Y2, .. ·,Ya > Yl = g(f3Y2' ... , [JY!)lfJ
since Y = :::1f3 is feasible whenever::: (= f3y) is feasible. The two preceding
inequalities do imply positive homogeneity, as was to be proved.
t The expression 'constant returns to scale' is explained as follows: if the first good is
the sole output, the return with respect to the input I in the productive transformation y' is,
by definition, the ratio yi!e- yn. This assumption specifies that the volume of output can
be changed without changing the return with respect to any of the inputs.
To characterise the second of the above assumptions, we often speak of
'non-increasing returns to scale' rather than of divisibility. The relationship
with the assumption of constant returns is obvious from the above formula-
tions. However, there must not be any confusion of the assunlQ!ion of
divisibility, or non-increasing returns to with the
shilJL be concerned.
-We 'also speak of non-decreasing returns to scale when f(yl) :s;; 0 (or
y
1
E Y) and a > 1 imply f(ay l) :s;; O.
Figure 6 illustrates the three situations for the case of a single input and a
single output. The production set bounded by r I relates to constant returns
to scale, that bounded by r Z to decreasing returns and that bounded by r 3
to increasing returns (of course, a given production set may come into none
of these three categories).
b = Y
I I
o
Fig, 6
CONVEXITY. If the vectors yl and yZ define two feasible productions and
if '0 < a < I: then the vector ayl + (I - a)yZ defines a feasible production.
In short, there is convexity if the set Y contains every segment joining two
of its points. Figures J and 2 correspond to the intersections of a convex set
Yof R4. Similarly, the sets in Figures 3, 4 and 5 satisfy the assumption of
convexity. Finally, in Figure 6, the set bounded by r
3
is not convex, and the
other two sets are.
Obviously divisibility and additivity imply convexity. Since the null vector
naturally belongs to Y, convexity implies divisibility (To show

56
The producer
Equilibrium for the firm in perfect competition 57
h = 2, ... , I,
this, we need only apply the property of convexity, taking the null vector
for y
2
.)
Convexity has consequences for the second derivatives of the production
function. To investigate these consequences, we shall deal with the case of a
function of the form
Here gj, is the value at yO of the first derivative of g with respect to y".
Similarly gi:k is the value at yO of the second derivative of g with respect to
y" and J'k. The two numbers e and 1] are infinitely small with the dy".
Subtracting (22) multiplied by CI. from (23), and taking account of the fact
that 0 < CI. < I, we have
1 1
I I dYh dYk :::; 0 (24)
h=2 k=2
4. Equilibrium for the firm in perfect competition
When dealing with the consumer, we reduced the problem of choosing
the best consumption complex to that of maximising a utility function. We
shall now assume that the firm tries to maximise the net value of its production:
I I I
PY = I PhI'" = I Ph
b
" - I Ph
a
". (25)
,,= I ,,= I ,,= I
To conclude our discussion, we return to the two reasons mentioned
earlier for departures from additivity and divisibility.
The fact that certain factors available in limited quantities have not been
taken into account explicitly in the formulation of the model obviously does
not affect the marginal returns to the other factors. On the other hand, this
fact may explain why we choose .functions for which returns to scale are
diminishing, while additivity implies constant returns.
The presence of considerable indivisibilities may explain the appearance
of production functions with increasing returns to scale for which the
assumption of non-increasing marginal returns is not satisfied.
M. Allais suggests that we distinguish two situations. In some branches of
production, divisibility can be considered to be approximately satisfied to
a sufficient extent. In this situation we usually find that production is carried
on by a relatively large number of technical units functioning in similar
conditions. The technology of this branch satisfies the assumption of constant
returns to scale. M. Allais uses the term 'differentiated sector' to cover all
productive activity of this ,kind.
In other fields, considerable indivisibilities exist. The market for each of
the goods produced is then served by a very small number of very large
technical units. To represent this situation, M. Allais assumes that a single
firm exists in each such field, all of which constitute what he calls the
'undifferentiated sector'.
This distinction will be taken again later, notably in Chapter 7 when we
shall consider economies involving a large number of agents.
that is,
0(- :::; O.
o( - Yh)
The marginal return to h(ogjoa
h
= - gj,), also called the marginal productivity,
is therefore a decreasing function of the quantity of input h used (ah = - Yh)'
We should point out that diminishing marginal returns and constant
returns to scale are not contradictory, as can be verified from the function
YI = -JY2Y3- imply both
scale and convexity, therefore non-increasing '*
-
(23)
YI = g(Y2, ... , YI)' (5)
Consider two infinitely close vectors yO and yO + dy which satisfy (5):
y? = ... , Y?) (19)
and
Y? + dYI = + dY2' ... , Y? + dYI)' (20)
If 0 < CI. < I, then yO + (X dy a possible vector; it therefore satisfies
Y? + (X dYl :::; + CI. dyz, ... , Y? + CI. dYI)' (21)
Let us assume that the second derivatives of g are continuous. Expanding
the right hand sides of (20) and (21) up to the second order, and taking
account of (19), we obtain
1 (1 + e) 1 I
dYI = I 0;, d.l\ + -2'- I I dy" dYk (22)
"=2 "=2 k=2
and
(the multiplier CI.(CI. - I + Cl.1] - e) IS certainly negative if the dy" are
sufficiently small).
Since a priori the dy" can have any values, convexity implies that the
IJIatrix G" of the second derivatives ghk i.)' negative definite or negative semi-
...
Conversely, it can be shown that, if G" is negative definite for any system of
values given to Y2, Y3, ... , YI, then the assumption of convexity holds.
The condition on G", which we have just established, is a general form of
the assumption of non-increasing marginqlrc!urns. In particular, this condition
implies
58 The producer Equilibrium for the firm in perfect competition
59
t Concerning the difficulties faced by the theory of management of firms and the
references dealing with it, see H. Leibenstein, 'The Mlssmg Lmk: Micro-MIcro Theory,
Jot/mal oj Economic Literature, June 1979.
inadequate for building a 'theory of the firm' that could serve as a gene.ral
conceptual framework for the discussion of the many problems concernmg
decisions to be taken by business managers. We must remember that the
microeconomic representation considered here aims at a theory of prices
and resources aIlocation not at a theory of the management of the
Adopting the assumptions of profit maximisation and perfect competitIOn,
and using a production function representing the technical constraints, we
can easily determine equilibrium for the firm. We need only maXImIse py
subject to the constraint
f(.1'I' .1'2' .. , )'J) = O. (26)
(In what follows, we assume that no price p" is negative, so the firm. loses
nothing by limiting itself to technically efficient net productIOns. ObvIOusly
we also assume that the price vector is not identically zero.)
If we follow the same approach as for consumption theory, we should no.w
investigate the existence and uniqueness of equilibrium. We shall not do thIS,
which in any case raises some difficulties of principle (see the footnote at the
start of Section 6). So we shall go straight on to consider the margmal
equalities satisfied in the equilibrium. .,
Maximisation of (25) subject to the constraint (26) IS a sImple case ot the
classical problem of constrained maximisation. The necessary first order
conditions for a vector yO to be a solution imply the existence of a Lagrange
multiplier Asuch that
h=I,2, ... ,1 (27)
where f(, is the value at yO of the derivative of f with respect to Yh' For the
application of theorem VI of the Appendix, it is assumed here that the fh
are not all simultaneously zero. It foIlows from the remark at the end of
Section 1 that the ff< are not negative and consequently that Ais positive.
Conditions (27) imply
This expression, which is the amount by which the value of outputs
exceeds the value of inputs also defines the 'profit' that the firm derives
from production. In fact, the microeconomic theory with which we are
concerned considers the behaviour of the firm to be motivated by its desire to
realise the greatest possible profit subject to the constraints imposed by
technology and the institutional environment. This assumption, adopted in
all theories of general equilibrium, has been subject to criticism. However,
no alternative has so far been suggested which stands up to examination and
can provide the basis for a general theory.t Also, some criticisms arise from
misunderstanding of the wide generality of the model under study. In order
to avoid the same errors, we shaIl later discuss the definition of 'profit' when
time and uncertainty are taken into account. For our present purposes it is
sufficient that the assumption of profit maximisation seems to afford the best
way for a simple systematisation of the behaviour of firms.
Again, we consider the firm to be in a situation of perfect competition if:
- the price of each good is perfectly defined and exogenous for the firm,
and therefore independent of its production decisions;
- and if, at this price, the firm can acquire any quantity it requires of a
good, or dispose of any quantity it has produced.
Of course, this is an abstract model of real situations. BasicaIly, it assumes
that the firm is small relative to the market, so that its actions have no
influence on prices. Moreover, it assumes that the demands and supplies
emanating from other agents are completely flexible so that they can react
instantly to any supply or any demand emanating from the particular firm.
This model is clearly inappropriate to the 'undifferentiated sector'. At the
end of this chapter we shall discuss the case of the firm in a monopolistic
situation and in Chapter 6 we shall briefly consider the formulations proposed
for other situations of imperfect competition. When in Chapters 10 and
11, we shall have explicitly introduced time and uncertainties, we shall
also understand that strictly speaking perfect competition implies a much
richer market system than the one actually prevailing.
Thus, the hypotheses of profit maximisation and perfect competition
have the advantage of being simple, but they lead to an idealisation that
may look strong with respect to an essentially complex reality. I repeat
that these hypotheses are introduced here in order to permit the building
of a general equilibrium theory and that, for this purpose, they may
provide an admissible first approximation. They would on the contrary be
t r must, however, mention here the existence of a general equilibrium theory for
economies with labour managed firms. The objective of the fi,nn is then said to be
maximisation of value added per worker rather than maximisation of profit. On this subject
see 1. Dreze, 'Some Theory of Labor Management and Participation', Econometrica,
November 1976.
Ps
f; Pr
In the equilibrium, the marginal rate of substitution between the.
modities rand s must equal the ratio of the prices of these commodltles.
In particular, if the production function is
YI = g(Y2' Y3' ... , YI),
conditions (27) become
PI = A and Ph = - for h f:. 1,
(28)
com-II
(29)
60 The producer
The case of additional constraints 61
(32)
(34)
- dt)yO
and so
h=2,3, ... ,l. (30)
P1
The marginal productivity of commodity h must equal the ratio between its
price and that of the output.
As in consumption theory, we can find the necessary. second order condi-
tions for a profit maximum. With the general form of the production function,
(26) say, these conditions require
/
I dYh dYk 0 (31)
h,k= 1
for every set of dYIl such that
/
I dYh = 0,
11=1
where, of course, J,;'k denotes the value at yO of the secon.d derivative of f
with respect to YII and Yk (see theorem VIII in the Appendix).
In the particular case of the production function (29), the second order
conditions imply more simply that
/
I dYIl dYk 0 (33)
lI.k = 2
for every set of dYh's (where h = 2, 3, ... , I). For, we can always associate
I
with these dYh's a number dY1 such that (32) is satisfied; (33) then follows
1
from So we come back to the assumptio.n ofnon-increasing marginal returns,
whzch zs therefore at an equilibrium for the firm.
order conditions reveal aE important point: the firm cannot
the PIQcLuction st:Lwhere reLl.![!}§"
!o scale are locally increasing. Let us take the case of the production function
(29) and inputs are increased by the quantities do:,
... , Y? do:. Let dY1 be the corresponding increase in output. We can say that
the returns to scale are locally increasing if dyddo: is an increasing function
of do:. If we consider a limited expansion of dY1 and ignore the case where
the second order term is zero, we see that the multiplier of do: in the expression
for dyddo: is
/
" ,,0 0
L, gllkYhYk'
h,k=Z
It cannot be positi vewithout contradicting the necessary second order condition.
Thus competitive equilibrium is incompatible with such increasing returns
to scale, which are often characteristic of the sector in which very large
production units predominate. The maintenance of equilibrium for this
sector demands forms of institutional organisation other than perfect
competition (see, for example, the case of monopoly in Section 9 below, or
the management rule for certain public services given in Chapter 6, Section 6).
We can also now consider the inverse problem and
conditions (27) are sufficient for an e"quilil:lrTum-or"the-ftrD; ifilie assumption
"aTeonvexlty is·satiSfiecl.TIieT6110Wiii"gpropertYlliemore ma-tcnesproposi tion
2"ffi<::I1apter:2;-relating" to the consumer. But its proof is much shorter.
lJ If the technical constraints are represented by a dif-
ferentiable production function defining a convex set Y and if the vector yO
satisfies (26) and (27) with an appropriate positive number l, then yO is an
equilibrium for the firm.
Consider a vector y
1
that is technically possible, but apart from that may
be any vector:
f(y
1
) O.
Let dt be a small positive number. Because Y is convex, the vector (1
+ dty
1
is technically possible, and so
f[yO + (y1 - yO) dt] O.
But f(yO) = 0, hence:
f[yO + (i - yO) dt] - f(yO) s; 0
-
If dt tends to zero, this inequality holds in the limit, and consequently
I
L - 0, (35)
h=l
where fh is the value at yO of the derivative offwith respect to Yh'
In view of (27), and since l is positive, (35) implies
I
L - 0.
11=1
The profit associated with y1 cannot exceed the profit associated with yO,
which is the required result.
5. The case of additional constraints
We have seen that the production function may be insufficient for complete
representation of technical constraints. Without going into details, we shall
discuss briefly the treatment of cases where additional constraints must be
added.
Suppose first that the constraints are represented by the production function
(26) and a second condition:
</J(Yl> yz .. , y/) = ° (36)
62 The producer
The case oj additional constraints 63
After introduction of a second Lagrange multiplier, the first order conditions
become:
t The introduction of such a composite good raises no difficulty when we are considering
the firm in isolation; but it is usually inappropriate for the discussion of general equili-
brium, since goods 3 and 4 may be produced by two distinct firms, or consumed by other
agents in a proportion other than <x.
P1Yl + pzYz + P3Y3 + P4Y4;
the constraints are
{
- fey!> Yz, Y3, Y4) 0
- Y4 0
Y4 - aY3 O.
----
Ph = + h = 1,2, ... , I, (37)
which replaces (27).
Does such a substitution have much effect on our results? Not necessarily.
A relatively simple alteration in the properties is sufficient in some cases.
Let us return to the example of four goods and the additional constraint
Y4 = aY3, (38)
which expresses strict proportionality between two inputs. System (37)
becomes
(41)
(43)
(42)
B
h ,= 1,2
h = 1,2.
Fig. 7
/
I
I
I
I
I
I
I
I
I
I I
I I
I I
I
I
I
I
o
Let A, fl.l and fl.2 be the corresponding Kuhn-Tucker multipliers. The necessary
conditions for a maximum are
for
{
Ph =
P3 = Ai; + afl.2
P4 = + fl.l - fl.2
where each of the multipliers A, fl.l and fl2 must be non-negative, and must be
zero when the corresponding constraint is a strict inequality.
If PI or P2 is positive, as we shall assume, the multiplier Amust be positive
and the equilibrium yO must strictly satisfy f(yO) = O. We can then dis-
tinguish three cases:
(i) If the equilibrium is such that 0 < - < - ayg (the point M on
Figure 7), the multipliers fl.l and fl2 are zero. System (41) reduces to system
(27) exactly as if the constraints (40) did not exist.
(ii) If the equilibri um is such that = 0 and yg < 0 (point B on Figure 7),
/(2 = 0 and fl.l O. After elimination of fl.l, system (27) is replaced by
{
Ph = h = 1,2,3.
P4
In particular, if the production function takes the form (5), the marginal
productivity - 94 of good 4 is less than or at most equal to the price ratio
P4/Pl'
(iii) If the equilibrium is such that = < 0 (point A in Figure 7),
/(1 = 0 and flz O. System (27) becomes
{
Ph =
P3 + ap4 = AU; +
P3 Aj; and P4
(39)
(40)
h = 1,2.
h = 1,2. for
{
Ph = Aj;,
P3 = Aj; - fla
P4 = + fl
Eliminating fl, we obtain
{
Ph =
P3 + ap4 = AU; +
This new system has the same form as (27) provided that goods 3 and 4 are
replaced by a composite good one unit of which consists of one unit of good
3 and a times one unit of good 4;/3 + af4 is then the partial derivative off
with respect to the composite good. t
Similarly, no insurmountable problem arises if we take account of con-
straints expressed by inequalities. Suppose, for example, that there are
again four goods and, apart from the production function, the two con-
straints
o - Y4 - aY3'
(Goods 3 and 4 are inputs, and the proportion of 4 with respect to 3 is
bounded above; see Figure 7.)
Here we have a case for application of theorem XI of the Appendix.
The function to be maximised is
64 The producer
Supply and demand laws for the firm
65
(47)
(45)
(44)
h = 1,2, ... , l,
h, k = 1,2, ... , t.
[
F" I'J_1
[f']' 0
while the right hand side is the element on the kth .and the hth colum,n.
Now, the matrix (47), which we assume here to eXIst, IS clearly symmetnc,
which proves the equality. d
This property shows that we can say unambiguously whether two goo s
are substitutes or complements for the particular firm.. We only at
the sign of the partial derivative Ol]h/OPk' More tat. :wo
t
' ts h al1d k are complements if thiS denvatlve IS 'pOSItIve,
outputs or wo mpu .
and are substitutes if it is negatIve.
su I unction is homo eneous 0 de ree zero with res ec.t .to
P
and for any multiplication of these prices by the same pOSItIve
PuPz, ... , I h t' t E Y orf(y) 0
hIS is an 0 vious property since t e cons ram, y ,
involve p and the function to be maximised is II1 p.
If yO maximises py subject to the' constraint, it also maXimIses rJ.py when rJ.
is positive. . . f I f ctions
Just as in consumption theory, this homogeneity 0 . supp Y
shows that the choice of numeraire doe.s no.t eqUIhbnum. Agam It can
be described as 'the absence of money IllusIOn. . .
oThe substitution eU!ct of h for k is equal to the
fpr h Consider the increase in the supply of h when the pnce 0 Iml.I1IS
the net supply functions are we can :haractense thiS
'substitution effect' of h for k by the pa:tial of I] h With respect to Pk'
Property (ii) then expresses the followmg equalIty:
Ol]h Ol]k
0Pk 0Ph
To establish this property, we differentiate the system consisting of (27)
and (26) and obtain
\
Art dYk + dA = dph
I = 0,
h= 1
which can be written in matrix form:
[
AF" I'J [dYJ = [dPJ, (46)
[fT 0 dA 0
with the obvious notation. This equality shows that the left hand side of (44)
is the element on the hth row and kth column of
This brings us back to (39); we can introduce a composite good for the
interpretation of the last equality; but we can now identify the irtdividual
marginal productivities of inputs 3 and 4 with respect to output l, namely
fi/f; and f;.!({. We see that the marginal productivity of input 3 is at most
P3/P1> and that of input 4 is at least P4/Pl' In fact, to incrc:ase the input of
factor 3 without changing the input of factor 4 is possible but not worth
while, whereas to increase the input of factor 4 without changing the input
of factor 3 might be worth while but is impossible.
In short, consideration of additional constraints entails some modification
in the equilibrium conditions but makes no basic change in their nature.
6. Supply and demand laws for the firm
The theory of the firm must lead to some general properties of supply
and demand functions, as happened with the theory of the consumer. In the
context of the perfect competition model, the supply function for com-
modity h defines how the firm's output of this good varies as the prices of
all goods vary. Similarly, the demand function for commodity h defines how
the firm's input of this commodity varies. We shall deal with these two
functions simultaneously by considering net supply, which, by definition,
is equal to supply for an output and to demand with a change of sign for an
input.
The net supply law for commodity h is therefore that law which defines Yh
as a function of the P1> P2, ... , PI' the set Y of feasible productions, or the
production function f, being fixed. We shall write this law rth(P1, Pz, ... , PI),
assuming that yO exists, and is unique, for every vector P belonging to an
I-dimensional domain of R'.t We can easily establish the following three
t In fact, this assumption is more restrictive than appears at first sight. For example,
if the production function satisfies the assumption of constant returns to scale and is
expressed in the form (5) or (29), the derivatives g{, are homogeneous of degree zero and
can therefore be expressed' as functions of the 1- 2 variables YZ/Y" .. " , Y,-I!Y,. Now,
there are I - 1 equations (30), necessary for equilibrium and also sufficient in the case of
convexity. If the Ph are chosen freely, these equations will not generally have a solution.
In the particular case where the Ph are such that a solution exists, yO say, then every pro-
portional vector exyO will also be a solution (ex > 0).
In economic terms these formal difficulties have the following significance. The decision
to produce can be split into two stages: (i) the choice of the technical coefficients Yl /Y" .. "
Y,-l/Y" (ii) the determination of the volume of production, In the case of constant returns
to scale, the two stages are independent of each other and, once the best technical coefficients
are chosen, profit is proportional to the volume of production. If it is positive, no equilibrium
exists since it is always advantageous to increase production. If it is negative, only zero
production gives an equilibrium which does not obey the marginal equalities (30). If profit
is zero, then any level of production is optimal.
The most modern versions of microeconomic theory take account of these difficulties:
net SUR-pIx. functions can be defined only for a subset of the values that are a priori
the
functions' is used.
66 The producer Cost functions 67
This is the general form of the relation of comparative statics, which must
be obeyed in the comparison of two different equilIbna for the same
firm.
In particular, if pI and pZ are identical except where price P" is concerned,
the inequality becomes:
( p ~ - p t ) ( y ~ - yt) ~ O.
This establishes property (iii).
® When the price of a good increases, the net supply of this good cannot'll
diminish. For the proof of this property we can use the second order condition
for an equilibrium and establish that the partial derivative of/lh with respect to
Ph is not negative. The reasoning is similar to that used for consumer demand
(cf. property 3 in Chapter 2, Section 9). We can also proceed directly on the
basis of finite differences, which makes the result clearer and more general.
Consider two price vectors, pI and pZ say, and two corresponding equilibria,
yl and yZ. Since yl maximises ply in the set of the feasible y's and since yZ is
feasible, we can write
the cost function changes when these prices change. The production set or
production function are more fundamental since they represent the technical
constraints independently of the price system.
In the second place, a production theory based on the analysis of costs is
out of place in a general equilibrium theory which treats prices as endogenous
and not determined apriori. Since our aim is to lead up to the study of general
equilibrium, we must start with production sets or functions.
However, an examination of cost functions reveals certain useful classical
properties which are simple to establish at this point and may be needed
later. We assume here that the markets for inputs are competitive so that
the Ph are given for the firm (h = 2,3, ... , I).
Since we restrict ourselves to the case of only one output, we can take the
production function as
YI = g(yz, Y3, ... , Yl)' (29)
Before defining the cost function, we must first find the combination of
inputs which allows production of a given quantity Y1 of commodity 1 at
minimum cost, so we must maximise profit subject to the constraint that
Yl = YI' This is a particular case of the problem discussed at the start of
Section 5 where ¢(y) = YI - YI' Here the system of first order conditions
(37) becomes
{
PI = A + ~
Ph = - Agh for h = 2,3, ... , [.
The first equation allows us to find /1 and is of no further use. If, as we
assume here, the first order conditions are sufficient for cost minimisation, the
solution is obtained by determining values of Aand of yz, Y3' ... , YI which satisfy
(52)
{
g(y:: Y3' ...,' yz) = ~ l
Ph - - Agh h - 2,3, ... , l.
(48)
(49)
(50)
plyZ ~ plyl
and also
pZyl ~ pZyz
or equivalently,
_ pZyz ~ _ pZyl.
Adding (48) and (49), we obtain
(pi _ pZ)yZ ~ (pI _ pZ)yl
or
:* (pI _ pZ)(yl _ yZ) ~ O.
We need only replace the Yh in this expression by their values in the
solution of (52) when we want to determine the cost function, which relates
the value of the minimum of C with the production level YI (the Ph being
When the firm minimises its cost of production, the marginal rates of
substitution of inputs are equal to the ratios of their prices; but the marginal
productivity of an input, hfor example, is not necessarily equal to Ph/Pl' It is
equal to P,.!PI ifYI is the optimal production for the firm selling on a competi-
tive market. But for freely chosen YI' in most cases it is not equal to this ratio.
Cost C is defined as
7. Cost functions
Suppose that the prices Ph of the different commodities are given and that
the firm produces only one good, the good I to fix ideas. The cost function
relates to the quantity produced YI, the minimum value of the input mix which
yields this production.
The theory of the firm is often built up on the initial basis of the cost
function. This greatly simplifies the analysis, but is subject to criticism on
two counts.
In the first place, the relationship between the value of input complex and
the quantity produced depends on the prices Ph of the different inputs, so that
I I
C = L Phah = - L PhYh'
h=Z h=Z
(53)
68 The producer Cost functions 69
hence, taking account of the definition of C and the marginal equalities (52),
C = AYl'
This equation, together with (55) shows that A, which a priori is a function of
Yb is in fact a constant (always assuming that the p" are fixed).t
t We saw that the assumption of constant returns to scale would usually not hold if
all the factors of production were not accounted for in the model. When defining marginal
cost, we assumed that the quantities of all the factors could be freely fixed. This latter
assumption is inappropriate to factors such as the work capacity of the managing director.
So the case of constant marginal cost is not necessarily frequent in relation to a firm some
of whose factors cannot vary. (See below the distinction between long-term and short-term
costs.)
(56)
(57)
(55)
h = 2,3, ... , I.
de = A L = AdYl'

This equation establishes that ). equals marginal cost.
We can also verify that the assumption of non-increasing marginal returns If
marginal cost is increasing or constant. Let us differentiate (52), \1
keepmg pnces constant:
r
t dy" = dYl

l + A ktz dYk = 0
Multiply the hth equation by dy,,; sum for h = 2,3, ... , I; take account of
the first equation: we obtain
I
dA dYl + A I dy" dYk = O.
z
Since marginal cost Ais positive, the assumption of non-increasing marginal
returns implies
dA
dA' dYl 0 or - 0, (58)
dYl
which is the required result.
So a cost curve derived from a production function with non-increasing
marginal returns is concave upwards. The classical curve of the. cost function, I.
as exhibited in Figure 8, is concave downwards at the start: thIS corresponds
to the range of values of output for which indivisibilities are significant and
marginal returns are increasing.
We note also that marginal cost is rigorously constant when the 1\
function satisfies the assumption of constant returns to scale. The functIOn g
is then homogeneous of the first degree, and so
I
I g;,Yh = YI;

considered as given). t This function is often assumed to have the form of the
curve C in Figure 8.
t The term 'cost function' is sometimes also used for the function that relates C to 11
and to Pl, PJ .. ,PI'
Fig. 8
When looking for the equilibrium of the firm, we can work in two stages:
(i) Define the cost function, that is, determine for each value of YI the
Yz, Y3, ... , Yl which minimise cost and find the value C corresponding to this
minimum cost.
(ii) Choose YI so as to maximise profit (pdl - C(YI))'
The solution of stage (ii) is obvious. The first order condition requires
PI = C'(jil)' (54)
C' measures the increase in cost resulting from a small increase in production,
and is therefore the 'marginal cost'. Equation (54) shows that, in competi-
tive equilibrium, marginal cost is equal to price of the output. The second order
condition requires that the second derivative of the profit is negative or zero,
that is, that marginal cost is increasing or constant.
We shall verify that, in (52), A equals the marginal cost. When marginal
cost is equated to price PI' the first order conditions for cost minimisation,
equations (52), are transformed into first order conditions for profit maximisa-
tion, equations (29) and (30).
Let us differentiate (53), the expression for cost, keeping prices p" constant:
I
dC = - L p" dy"

or, taking account of (52) and, in particular, differentiating the first equation,
70
The producer
Short and long-run decisions 71
Fig. 9
(iii) the firm should increase production ..price PI)' .
As we said previously, the existence of situatIOns (I) and together WIth
the multiplicity of equilibria in (ii), are sufficiently real possIbIlItIes to us
avoid trying to prove for producer equilibrium a general property
and uniqueness corresponding to that stated for consumer eqUIlIbnum 111
proposition I of Chapter 2.
8. Short and long-run decisions
Cost minimisation has just been presented as a stage in profit
In fact abandoning the strict model of perfect competition, we sometImes
that some firms actually behave so as to provide an exogenously
determined output and minimise their production cost. System (52) then
applies directly to the equilibrium for the firm.
Similarly, in some contexts, the firm does not choose all, but only some of
its inputs, the others being predetermined. Thus for the s.ame firm. we. often
distinguish between long-run decisions relating to the entIre orgamsatlOn of
production (choice of equipment and manufacturing processes) and sh?rt-run
decisions relating to the use of an already existing productive capacIty. So
for short-run decisions, the inputs relating to capital equipment are fixed.
situations can easily be analysed using the principles applied above.
Suppose, to fix ideas, that capital equipment is represented by a. s.ingle gO?d,
the /th. Let YI be the predetermined value ofYI' The short-run deCISIOn conSIsts
of profit maximisation subject the constraint YI = YI' The cost
function relates cost C to the Yl of output :vhen Y.I = YI' _
inputs Y being fixed so as to m1l1ImISe cost. Let thiS functIon be C (y I, YI).
As before, we see that inputs Y2' Y3' ... , YI_I> cost C* and marginal cost),*
obey the system
c

o
,7
,
,

"
, 1
./ i
, 1
" I
em
"
" ....

In addition to total cost C and marginal cost C' we often consider average
cost per unit of output, namely c = C/h. If we differentiate c with respect
to YI, it is immediately obvious t'hat average cost is increasing or decreasing
according as it is greater or less than marginal cost (a typical curve c appears
in Figure 8).
It is sometimes convenient to give a diagram representing the last stage in
profit maximisation. Let the curves c and y represent respectively variations
in average cost and marginal cost as a function of YI for given values of
Pl, P3, ... , Pi- The equilibrium point yO is determined by the abscissa Y? of
the point on the curve }' whose ordinate is PI' The profit is then Y? times
the difference in the ordinates of the points on y and c with abscissa Y?
Examination of the figure rounds off the preceding analysis, which was
limited to finding necessary conditions for a profit maximum at a point yO
for which constraints other than the production function do not operate.
Are these conditions also sufficient, as we assumed earlier when we said that
y? corresponds to the equilibrium?
Ambiguity may exist if several points on y have PI as ordinate. In practice,
this is likely to arise only in two ways. In the first place, there may be two
such points, one on the decreasing part and the other on the increasing part
of the marginal cost curve; the first point cannot correspond to an equilibrium
since it does not satisfy the second order condition, so that the ambiguity
disappears. Also, at the ordinate PI the curve y may be flat (in particular,
we saw that marginal cost is constant if the productwn function satisfies the
assumption of constant returns); all the points on this flat section give the
same profit; if one of them corresponds to an equilibrium, then the others
also correspond to equilibria.
The point or points with ordinate PI and lying on the non-decreasing
part of y may not correspond to an equilibrium if it is to the interest of the
firm to have zero output YI' This situation arises ifPI is less than the minimum
average cost Cm and if Yl = 0 implies zero profit, since the points considered
then give negative profit.
Finally, if the whole curve y lies below the ordinate corresponding to PI>
there is no limit on the increase of profit and it is to the interest of the firm
to go on increasing production indefinitely. (Of course, in practice it would
come up against a limit sooner or later, but the chosen cost function ignores
this fact.)
To sum up, for given values of Pl, P3, ""PI' the value of PI may be such
that:
(i) the firm should choose YI = 0 (low price PI);
(ii) the firm should choose a finite output Y?, which mayor may not be
defined uniquely;
long-run cost, gives the value YI for J'l. For, the solution of (52) then satisfies
(59) with c* = C. Let yy be this particular value of YI' At yy, the equality
PI = - A*g; is satisfied, so that dC* = A* dYI = dC. At this point, long and
short-run marginal costs are equal, long and short-run average costs are
tangential. A priori, this may seem an obvious result, since if existing equip-
ment coincides with what the firm would choose in the long run in the same
73
(61 )
Monopoly
price situation, then short and long-run equilibria must naturally coincide.
Hence, the long-run average cost curve is the envelope of short-run
average cost curves (obviously the same property holds for total cost curves).
In any case, the short-run cost cannot be lower than the long-run cost smce
the minimisation which defines the former is subject to one more constramt
than that which defines the latter.
t The assumption of independence of demand with respect to prices Pz, ... , p, is made
here for the sake of simplicity. It can obviously be eliminated if prices Pz, ... , p, are
independent of the decisions of the firm, that is, if the markets for all goods except the first
are competitive.
9. Monopoly
The formal approach developed so far is more or less easily transposed to
institutional situations that differ from perfect competition. We may briefly
examine here the classical theory of monopoly, leaving for Chapters 6 and 8
the analysis of other situations.
In the applied study of market structures a firm is said to have a monopoly
position on the market for commodity h if it supplies alone this commodity
and if demand comes from many agents who are individually small and act
independently of one another. Classical monopoly theory represents this
situation starting from the hypothesis that the same price Ph will apply to the
exchange of all units of commodity h but that this price will depend on the
quantity )'h that the seller will supply. Thus the monopoly faces a demand
whose quantity varies with the price of his product but is otherwise inde-
pendent of his decision.
The firm facing such a situation necessarily takes account of the fact that
the price at which it will dispose of its output depends on the quantity which
it puts on the market. We can no longer analyse its behaviour on the
assumption that it considers price as exogenous. We have to adopt a formal
model other than that of perfect competition.
Suppose, for example, that the firm produces good I and sells it on a market
where there are many buyers whose demand depends on price PI and not on
other prices. t We can represent this demand by a relation between PI and YI:
where nI is the function defining the price at which the monopolist can
dispose of the volume of production YI'
lt may also happen that a firm is the only one to use a factor h (for example,
when it is the only employer of labour in a town). lt is said to be in a situation
of 'monopsony'. It knows that price Ph depends on the quantity a
h
= - Yh
(59)
eL
yO yC yL
, I ,
Fig. 10
o
The producer
I
g(Yz, ... , YI-I, YI) = h
Ph = - A * g ~ h = 2, 3, ... , I - 1,
1-1
I
C* = - L PhYh - PIYI'
h=2
72
Differentiating the first and last equations for given Ph and taking account
of the intermediate equalities, we obtain
dC* = A* dh - (A*g; + PI) dYI,
which replaces (55). The short-run marginal cost is again equal to the
equilibrium value of the Lagrange multiplier A*. We could also verify that,
to determine the value of YI which maximises profit subject to the constraint
YI = YI' we must add to (59) the condition that the marginal cost A* equals Pl'
Let us illustrate this theory by a diagram in which the different cost
functions are represented as a function of YI' Let cL and yL be the long-run
average and marginal cost curves. The long-run equilibrium value ofproduc-
tion for price PI is determined as the abscissa yf of the point on yL whose
ordinate is Pl' Also let cC and yC be the short-run average and marginal cost
curves. The short-run equilibrium is determined by the abscissa yC of the
. I
pomt on yC whose ordinate is Pl'
The long and short-run average cost curves generally have a common point
corresponding to the value of YI for which the solution of (52), defining the
TC n
eC
I
.' /
, .
~ - - - - - - ~ - - - - - - - - - t ~ 1
.'/'
II ,
y;
);
'. I
/1
I
I ,
I I
I I
I ,
I I
I i
Maximisation of, expression subject to the constraint expressed by the
productIOn functIOn Implies the following first order conditions:
75
(67)
Monopoly
We could apply the same reasoning to the case of pure monopoly where all
the 8
h
except 8
1
are zero. However we shall adopt a rather different approach
for an alternative presentation of the analysis, which is thus reinforced.
As in the case of perfect competition, we can maximise profit by means of a
two-stage procedure involving first cost minimisation and determination of
the cost function. For a pure monopoly, cost minimisation is carried out in
exactly the same way as for a perfectly competitive firm and the cost ft:nction
is exactly the same. So we can confine ourselves to the second stage, and
find the value of Y1 which maximises
1t1(Y1) . Y1 - C(Y1)'
We can write this expression in its usual form
R(Y1) - C(Y1),
provided that 8
1
i= - 1 in the equilibrium, which we assume for simplicity.
The marginal productivity of the factor h is no longer equal to the ratio of
prices but to this ratio multiplied by a term depending on the elasticities
relating to the factor h and to output.
Consider first the case of a monopsony for which all the 8" are zero except
that relating to a particular input k. Equations (66) then reduce to the perfect
competition equations except for the kth, where - gk must equal pdp1
multiplied by the term (1 + 8k) which is usually greater than 1. The equili-
brium is therefore the same as in a situation of perfect competition involving
the same prices for all the goods except k, whose price is greater than that
actually asked by suppliers. Since, in the competitive situation, the firm's
demand 1] k can only decrease, the firm in a position of monopsony usually
employs a smaller quantity of the factor k than it would employ in competi-
tion. For this reason it may be said to be in the interest of the monopsonist to
adopt a 'Malthusian policy'.
where R(Yl) denotes the firm's receipts from output Y1'
Profit maximisation implies that Y1 is so chosen that
R'(Y1) = C'(Y1) (68)
and
R"(Y1) C"(Y1)' (69)
Equation (68) generalises condition (54) obtained for the case of perfect
competition.
We can easily compare monopoly equilibrium with equilibrium for the
firm in perfect competition. Figure 11 shows the average cost and marginal
cost curves c and y, as well as the curve d representing the demand function
1t
1
(Y1), that is, average revenue, and the curve (j representing marginal
revenue, that is, the function 1tl + Yl1ti. Suppose that 1ti is negative, as will
(63)
(29)
(66)
The producer
h = 2, ,.. , I
h = 1,2, ... , I,
-
74
which it uses as input. If it takes no account of the possible interdependence
of Ph and the pnces of other goods, the firm will fix its decisions as a function
of a supply law
p" = n,,(y,,) (62)
representing behaviour of the agents supplying the factor h and indicating
the pnce Ph whIch the firm must pay to acquire a quantity - y" of h.
. We, note that the case of perfect competition corresponds to the particular
sItuatIOn where n [ and nh are constant functions. Therefore we can deal
sImultaneously with monopoly and with monopsonies concerning one or
more factors by treating the case where the firm tries to maximise its profit
and takes account of functions n
h
relating the price of each good h to its net
production Yh (h = I, 2, .,., I).
As a function of y the profit, or net value of production, is
I
I n,,(y,,) , y",
h= [
where nhis the derivative of nil and }, is a Lagrange multiplier.
For w.hat follows, we shall consider the case where prices are non-zero and
shall WrIte the above conditions in the form
p,,(1 + 8,,) = h = 1,2, ... , I, (64)
taking account of the fact that Ph is the value of the function n and defining
81, as the inverse of the elasticity of demand (or supply) whichhoccurs in the
for. the good h because of agents other than the particular firm under
consIderatIOn:
d log nh
8" = y";,, = d log /y,J (65)
In,the case of perfect competition, market demand and supply are perfectly
elastIC from the standpoint the firm; the 8
h
are zero. Conditions (64)
reduce to the first order condItIOns (27) obtained earlier.
In ?rder to investigate (64), we shall consider the case where the production
functIOn takes the form
Y1 = g(YZ'Y3' ""Yl),
the good 1 being the firm's output. Equations (64) imply
p,,(1 + 8
h
)
Pl(1 + 8
1
)
76 The producer
Monopoly 77
necessarily be the case except perhaps for an inferior good; (j then lies below
d. According to (68), monopoly equilibrium is determined by the abscissa y!
of the point of intersection of y and (j. If the firm behaves as in perfect
competition, that is, if it takes no account of the reaction of price PI to its
supply YI, the equilibrium point is determined by the abscissa y? of the point
of intersection of y and d.
The study of monopoly has taken us outside the field of perfect
competition. We shall not pursue this line for the moment, but shall take
it up again in Chapters 6 and 7. However, two remarks may usefully be
made already at this stage.
In the first place, it is clear that situations of imperfect competition may
involve consumers as well as firms. For example, it is conceivable that a
t We should also note that, for the definition of the cost function, second order
conditions implying concavity of the isoquants in the neighbourhood of the equilibrium
must be satisfied. When this is not so, no equilibrium exists as long as the markets for the
factors are competitive: but a monopsony for the firm may allow equilibrium to be realised.
c
Fig. 12
' ....
........ "
........... _ "I
....... _- ',!
--- __, I
T
I ...
i ........
I
I
I
o
particularly wealthy consumer may have such influence on a market that he
has a position of near-monopsony.
In the second place, the theory of imperfect competition cannot depend
entirely on the constrained maximum techniques which we have used up
till now.
Of course, situations other than those we have considered can be dealt
with by constrained maximum techniques, for example, the case of a firm
that has a monopoly on each of the two or more independent markets in
which its output can be sold. In most cases, profit maximisation leads to
price differentiation, the firm releasing to each market a quantity of its
product such that marginal revenue from each market equals its marginal
cost over all its output.
Generally we can say that constrained maximisation is appropriate to the
extent that all agents except at most one adopt a passive attitude, taking the
decisions of other agents as given. This is just the situation for a monopoly,
since those who demand the proc;iuct accept as given the price which results
from the firm's decision on production. They have no other possible attitude
if their number is. large and they are all of the same relative importance, and
if they are unable to band together in opposition to the monopolist.
But imperfect competition is not limited to such situations. On some
,t'
I
I
I
/
I
I
I
C
I,
I ,
I "
: "b
1
I
y. yO
1 1
Fig. 11
o
At the point of intersection of y and d, the marginal cost must be non-
decreasing for y? to correspond to a true competitive equilibrium. It follows
from the fact that d is decreasing and from the positions of d and (j
that y! is necessarily smaller than y? The firm produces less in a position of
monopoly than in a situation of perfect competition involving the same prices
for it; this result is similar to that encountered earlier for monopsony.
We can consider R" as negative in the interpretation of (69) defining the
second order condition for a maximum. In particular it will be negative if
there is constant elasticity of demand, since then [;;1 is a fixed number, R' is
equal to n1(1 + [;;1) and R" to n'l(1 + [;;1)' The second order condition is
therefore satisfied for any situation where marginal cost is increasing.
But we should point out that this condition may also be satisfied in
situations where marginal cost is decreasing. More generally, monopoly may
sometimes allow an equilibrium to. be realised which is not possible in perfect
competition. Figure 12 shows an example for a firm with continually decreas-
ing marginal cost, which is possible in the "undifferentiated sector". t
p.
1
pO
1
78
markets there are relatively few buyers and sellers; on others, coalitions take
place. Other methods of analysis are necessary to deal with such cases.
We shall return to imperfect competition in Chapters 6 and 7 in order
to clarify problems of general economic equilibrium. We shall then see
how it relates to the theory of games.
4
Optimum theory
Up till now we have been considering the behaviour of a single agent.
With the theory of the optimum we approach the study of a whole society.
We therefore change our perspective and attack the problems raised by the
organisation of the simultaneous actions of all agents.
The classical approach would be first to discuss competitive equilibrium,
keeping to the positive standpoint of the previous lectures, and then to go on
to the normative standpoint of the search for the optimum. However, we
shall reverse the order of these two questions.
Optimum theory involves a rather simpler and more general model than
the model on which competitive equilibrium theory is based. It seems
plausible that the relationship of the two theories will be more clearly
understood if those assumptions which are not involved in optimum theory
are introduced in the later discussion of competitive equilibrium.
We are interested, therefore, in the problem of the best possible choice of
production and consumption in a given society. Clearly it may appear very
ambitious to attempt to deal with this. But it is one of the ultimate objectives
of economic science. Preoccupation with the optimum underlies many
propositions briefly stated by economists. By providing an initial formalisation
and by rigorously establishing conditions for the validity of classical proposi-
tions, optimum theory provides the logical foundation for a whole branch
of economics.
We must first find out what is meant by the 'best choice' for the ~ o c i e t y
and go on to study the characteristics of situations resulting from this choice.
1. Definition of optimal states
Before fixing a principle of choice, we must again define what are 'feasible'
states.
For our present investigation, a 'state of the economy' consists of m
consumption vectors Xi and n net production vectors Yj'

or possibly to the whole. This will be the case if Y is a closed set and if.. ThiS wIiI be dIscussed 111 Section 2. It docs not restrict the validity of the theory to assume this. without leaving Y. and such that /(1'[. Others like V. Later on we s a P01l1.~t~ further stress on it. . ome cxposltlon easier an .n consumption theory.0) the second by a hyperplane (YI . we can find an efficient y* such that )'/~ :> )'h for all h. One feature is the existence of 'proportionalities' or 'coefficients of production': some inputs must be combined in given proportions. I. Another feature is the possibility of substituting an input for another: machines can replace men. Economists such as K.. .' d to allow the free use of differential calculus. i. Le~tures such as the present ones should not ignore th. we shall not lay Since this has already been dlscusse sU 1Clen y I .J ! . ProductIOn functIOns must t ~Ie ore. respect . .. )'/) = 0 real function (2) if and only if Y is an efficient vector. in their approach to the problems of general equilibrium economists have alternatively used two types of formalisations. Y2.ts .Ie If ¢ IS a rea un I defined unique y. a pnon.We may point out that.Dejinitiol1s 46 47 The producer and substitutions.ultaneous:~ the substitutability of machines for men and the proportIOnality ~e.2 . of the boundary of Y in the commodity space. or fixed coeffiCIent d . Y2 . the vector yl is said to be technically efficient if it belongs to the set Y of feasible net productions and if there exists no other vector y2 of Y such that l y/~ :? y/. h h' represented by production functIOns.. The firm's technical experts must eliminate the former in favour of the latter. we cannot increase one component of y indefinitely without reducing another. . . Following a tradition of almost a century.Ulvalentl y todrepr th the technical constraints on productiont (the functIOn / depen s on e particular producer j. _ . corresponding to every y of Y. a defined on R 1 such that: It]'!. Relative to such a transformation.duction {unction (for a particular finn is. one fuel can be substituted for another. mathematical economl'sts often introduce another more restrictive concept.i can e essel1tJ . output of one good cannot be increased without increasing input or reducing output of another good. or examl. Y b 'al proofs will be given under the assumption that the se. . we have ~he cholce to~ay d between two types of more specific forma!is~~i~n: either pro uC~lOn functions.. ) 4 . This is why we can often confine ourselves a priori to technically efficient net productions. and 2 represent two intersections . fi d b For the moment we shall not inquire Into the condl~lOns to be ~a:ls e .Y/) ~ 0 (3) if and only if)' belongs to Y.2. f Whe. the first two of which are outputs of the firm and the last two Il1puts. and coal Hence the theory built directly on Y IS fully genera III t IS Iron ore. tg~esol. t t 111 passll1g qUIre h 11 deflned and discussed with some care.~:~~:~1~nS. even th. . . some may require greater inputs and yield smalfer outputs than others. tl .y Y if weare to be able to define such a function. .J?D2.r large s~~stitutabI!Jtles. ou th~se places where the use of such functions conceals some difficulty. that there are four commo lIes. as does Y).t sim. Formally. we can use (I) or (3) eq. more or less fertilizer can be put in a given piece of agricultural land and more or less labour can be spent on it. processes combined into 'actIVity analySIS models. complementarity between inputs. however. . which formalises in particular the idea that marginal substitutions between inputs are feasible. esent According to this definition. therefore. !:::.~~c~~oi~sl:::u~. by definition.. The definition of Y can take into accoun. . like tile Ulilit y] ffunccttiOonn" wthlethP. usually allowing fo.of Y. . which stress two opposing features of production. . that of the 'production function'. In fact it will be used extensively with ~he aIm of m~kll1g . t -t'. Actually. . for example. for h = 1.e.. By this we mean any transformation which cannot be altered so as to yield larger net production of one good without this resulting in smaller net production of some other good.e pro uctlOn function concept. This is the method adopted in the most modern approaches to the subject. tl~ ~:~t b Y \~r. Walras in the first editions of his treatise constructed their systems assuming fixed proportionalities. The great advantage of the modern set theoretic approach is to cover both complementarities Rigorously. and which is zero when Its argumcndt IS ~ro.tweh' . So the technically efficient vectors y belong to a subset. then ¢(f) corresponds to the same set as f. we can confine ourselves to technically efficient vectors only if. ". tlOn IS not re.t In the construction of optimum and equilibrium theories we could impose on ourselves to use the production set Yas the sole representation of technical constraints. .oug t I~ assump h' f b ' d for the validity of the result. . Pareto have used formalisations implying that substitutabilities are everywhere prevalent.1 we want to build models that lend themselves to compu~atlOn or dealing with questions of applied economics. Geometric illustrations of the production set and the productIOn fu~c~~on are often fruitful. e . Suppose. hence the same crop may be achieved with a little less fertilizer and a little more labour.p~:I~~~t [~e~ia~~ = . like iron ore and coal in the process of producing pig iron. Marx or L.··.) 4 t' tl set of tJv' productions that are feasible from the quantllJes rcpresen s 1e ~ . I F .)'2.

To avoid confusion. This expression is similar to that encountered in consumption theory. for s. . in passing that a set which. after an obvious change in notation. . f(yO + dy) can be positive only if fk is not negative.. to fix ideas.Y~ and a2 = . == ~ g.fur the producer in gueiliml. even if these are several outputs..Yf) = YI . since f(yO) is zero. 1 Fig.fh denotes the value at yO of the derivative of f with respect to Yh' In particular. "·. The most general form of a production function is that in (2).) The function g* will generally be increasing with the a h and thefunction g will consequently be decreasing with respect to the Yh' or at least non-increasing. . equalities of the type for s =f. .g(Y2' .. It is often called the marginal productivity of s. The ratio (II) measures the increase in production resulting from an increase of one unit in the input of s (note that Y s is equal to minus the input). g. then (8) reduces to (9) or dYr dy'. represents the technically effi~ient combinations of inputs yielding given quantities of outputs is called an zsoquant. dYr and dys' are zero.. . There should be no real possibility of confusion from this ambiguity. r =f. 1. . Thus b i ~ g(. (4) (5) and The technical constraint is * YI ~ g(Y2' . Thus it is often assumed that the firm has only one output. the production function is then given the form:t In the particular case where (10) become .. a marginal rate of substitution. Let yO and yO + dy be two neighbouring technically efficient vectors. f(yO + dy) is positive. 7~' f~ (10) c:o-r----------a3-=-_-~ Fig. . This is. the second represents the set of inputs allowing the quantities b? = Y? and bg = yg of the two outputs to be obtained. Since yO is technically efficient. Later on we shall often assume that the function f is twice differentiable. the additional quantity of input of r which is necessary to compensate in output for a reduction of one unit in the input of s. the good I. 2 The ratio on the right hand side of (10) can be called theJnarginal rate of b.arginal r{lte of tra~ substitution formation..a2. The points satisfying (2) are represented by the North-Easi boundary on Figure 1 and the South-West boundary on Figure 2. Yf) - dYr -~ dy'. that is.~LihLgQQds-L~lliLL..a J. .aa or. Here the function g representing a given set Y is determined uniquely. like the curve in Figure 2. if all the dYh except two. ryve note '. I (11) ~f(YI'Yz. yO + dy is not technically possible.48 The producer Definit ions 49 II a~ = .y2 of the two inputs. we shall sometimes speak instead of the !!!:. in most cases we can solve the equality fCy) = 0 for YI and so revert to (5). .y/). a z)· (7) t Obviously this particular form is no longer affected by the indeterminacy already mentioned in relation to the general form (2). The ratio (12) defines. aJ. apart from sign. In fact. But. We note also that the first derivativesfh of the production functionfmust take non-negative values at every technically efficient point yO. (12) and the expression 'production function' is also used for the function g which defines the output resulting from given quantities of inputs.g~ f takes the form (4).. in fact. (6) b l ~ g*(a 2. Note that we could show inputs and outputs explicitly in (5). We can write h=1 L / f~ dYh =0 (8) b2 • Y2 where.. Slightly more particular expressions are often used. Consider a small variation dy all of whose components are zero except dYk' which is assumed positive.

with the utility function (8) illustrated in Figure 7. { y" ~ for the same list of goods h. the domains of variation of all.. these consequences are ditDcult to accept. (ii) In the second place. if the vectors {y\. the set Y has boundaries corresponding to non-technically efficient )roductions (for example. (2) is replaced by f(Yl' Y2. Differentiability implies that f is continuous and consequently that Y is a . closed set in RI. there exists a production function/. . differentiability assumes that... For example. since . for Figure 3. the isoquants do not have the same form as in Figure 2. (In fact.) The real situation is sometimes less clear-cut than Figure 4 aS5umes./(y) should be equal to a negativ~ num?er.e y". (The situation is similar to that in Chapter 2. the first derivatives off with respect to Y3 and Y4 are not continuous. they look like the isoquant in Figure 4. in some productive operations the different goods which constitute inputs must be combined in fixed proportions. (13) and (14..} of a convergent sequence each define a feasible production then the limiting vector certainly corresponds in reality to a feasible production.) However. a 3 and a 4 must take values whose ratio corresponds to that defined by the half-line GA. (15) M and Fig. the marginal rates of substitution are all welldefined. ° (13) To characterise the set of technically efficient vectors.) Because of the limits on the domains of variation of SOll.. Except at the point A. We shall presently return to this point. That is. technology may demand that some good r occurs only as input and some other good s only as output. Taken literally. the half-line GN in Figure 3).YZ' . but this does not alter basically the nature of the results. with respect to any technically efficient vector. or some. (i) In the first place. the second inequality can be eliminated if we assume that the firm can always dispose of its surplus without cost. .50 The producer The validity of production functions 51 2.. (At a point such as N. since this assumption is naturally expressed as: yO E Y and y" ~ y~ for all h implies y E Y. Aswe-sawli1-consumption theory. y 2 . 3 Yl { yz + ~ o:Y2 = 0. So the inequalities Yr ~ and Ys ~ 0 appear in the definition of Y. y/) = 0. The validity of production functions We must now investigate the conditions to be satisfied by the production set Y in order that. ° (14) Thus. . we shall proceed as if the limits on the domains J2f variation of the. of the y" may be limited. •. become y\ { yz + ~ o:yz ~ 0. ° (16) Ysatisfiesf(y*) = 0 since it can be approached both by a sequence of vectors But the continuity off implies also that every point y* on the boundary of y such thatf(y) ~ 0 and by a sequence of vectors such thatf(y) > 0.. first of all. So the 1\ definition off implies that every point y* on the boundary of Y is technically efficient. Apart from the surplus of one of the two inputs. If there is free disposal of surplus. . 0. When such proportionality ratios exist. b~~ shoul~ be positive ~or ev~ry point n~ar !y whose second coordll1ate IS pOSItIve.y/) ~ 0. At the point A. the half-lines AN and AM correspond to ~ non-technically efficient productions. The existence of such boundaries is incompatible with the cOlltinuity of f together with the conditions thatf(y) < is satisfied for every non-technically ° ° This complication will not be taken into account in our discussion of the general theories. This property is not restrictive. and in the second place. we can take account of these limitations by altering the definition of the production function and explicitly adding inequalities to the formal representations of the set Y and the set of technically efficient vectors.l'~re neveriil force. This is particularly the case for most of the raw materials used in many industrial processes. For example. thIS is incompatIble WIth the contll1Ulty off at N. 2 I efficient vector of Yand thatf(y) > is satisfied for every vector y outside Y. Moreover. ~ertain new particular features are revealed if we take account of constraints expressed by inequalities. These conditions are certainly more restrictive than it would appear at first glance.. { Yh ~ for a specified list of goods h. that this function is differentiable. to characterise Y we replace (3) by ° [(Y\.

5 ADDITIVITY. Assumptions about production sets ~I/ ~--. since. to integrate into the theory formalisations which describe technical constraints more accurately than do production functions. For this reason. For example. In this case. 1958. McGraw-Hili. if. (18) The theory becomes very complicated if such constraints are taken into account. they are better ignored in a course of lectures whose aim is to provide the student with a sound grasp of the general logic of the theories to be discussed rather than the difficulties which are t It is the aim of a new branch of economic science. Knowledge of these results is essential for the student. For. The firm can employ the two t~chniques simultaneously to produce the same quantities of outputs. New York. For example. In order formally to represent such situations as those of Figures 4 and 5. Linear programming alld activity analysis. even if their application is somewhat restricted by the simplifications required to justify the production function. Additivity fails to hold only if yl and y2 cannot be applied simultaneously. Berkeley 1951. realisation of yl -+ y2 requires double the actually available quantity of land. Similarly. The changes in production theory introduced by their presence will be described briefly. it may happen that the model does not identify all the commodities which in fact occur as inputs in production operations. This appears a natural assumption. _!/ 3 ~IC-- . See also Dorfman. if the capacity for work of the head of an industrial firm does not appear among the commodities. University of California Press. as in Figure 4. then the vector y = )'1 -+ y2 defines a feasible production (therefore y E Y or fey) ~ 0). and if his capacity limits production. Samuelson and Solo'. as in Figure 5. to be defined presently) then the same output can be obtained by simultaneous use of the two techniques on this new scale. Moreover.52 The producer Assumptions about production sets 53 there may be available to the firm two or more production techniques each requiring fixed proportions of inputs. _ o Fig. then additi vity no longer strictly applies. if each technique can be employed on a scale reduced by one half relative to that represented by A or B (the assumption of constant returns to scale. A very good account of thc resulting modifications is given in Dorfman. we can add other constraints to the equation fey) = 0 to characterise the set of technically efficient vectors. this approach alone leads to certain results which every economist must know. it seems that we can always realise y by realising independently yl and y2. This is the approach adopted in the most modern treatments of the theories with which we are concerned here.Y4 ~ .aYJ. the first represented by the point A. we see that the above-mentioned difficulties can be avoided if we base our reasoning directly on the set Y of feasible productions and on the set of technically efficient productions rather than on the production function. 'activity analysis'. encountered in their rigorous exposition. but not at A nor at B. As when a utility function is substituted for a preordering of consumer choices. the proportions differing for the different techniques.". the first derivatives offare in fact continuous at each point within AB.~ 8 3 fA M We must now discuss certain assumptions which are frequently adopted about production sets or production functions. the substitution of a production function for a production set makes exposition easier since it allows the use of the differential calculus and of fairly standard types of mathematical reasoning. Application of Linear Programming to the Theory of the Firm. each point on AB defines a possible combination of the two techniques yielding the same output as A or B. there must be a fixed proportion between Y3 and Y4' we write: Y4 = aY3' (17) In the case of two techniques. the second by the point B. if the available land is totally used by yl on the one hand and by y2 on the other. However. the point on Figure 5 corresponding to this method of production is the midpoint of AB. 4 Similarly. A priori there seelTl$ no reason for this to be the case. t Finally. Fig. the supplementary constraints may be . then additivity does not apply to its production set. Figure 5 relates to an example of two techniques. if the land in the possession of an agricultural undertaking does not appear among the commodities. . 3. If the two vectors y 1 and y 2 define feasible productions (yl E Y and y2 E Y or f(y 1) ~ 0 andf(y2) ~ 0). For example.f3YJ ~ .

. Figures J and 2 correspond to the intersections of a convex set Yof R4.. on the one hand the hypothesis implies. as was to be proved. . Figure 6 illustrates the three situations for the case of a single input and a single output. convexity implies divisibility il1-.n o~ jncreasiDKmarqi!!gLreturn-s.. . k y l are feasible. Taken literally.1. However. (To show --.. 0) and if 0 < a < 1. (k . additivity shows that f3y l = (f3 . the same hypothesis implies: g(Y2. [JY!)lfJ since Y = :::1f3 is feasible whenever::: (= f3y) is feasible. finally. y/) for every vector Y and every positive number fJ.~~-----<--_.. CONVEXITY. Finally. the ratio yi!e .yn. Indeed.. Conversely. t If the vector yl defines a feasible production (yl E Y or f(yl) :s.. .. O. .- . 4 and 5 satisfy the assumption of convexity. . a given production set may come into none of these three categories). there must not be any confusion of the assunlQ!ion of divisibility. .#.ill apply the-property oTadditivit. g(f3yz.. ... The two preceding inequalities do imply positive homogeneity.. . t The expression 'constant returns to scale' is explained as follows: if the first good is the sole output. that bounded by r Z to decreasing returns and that bounded by r 3 to increasing returns (of course.k)yl + k y l is feasible. If the vectors yl and yZ define two feasible productions and if '0 < a < I: then the vector ayl + (I . divisibility shows that (f3 . F. .. positive number. In short. by definition. Consider the particular case where the technical constraints are expressed in the form (5). the sets in Figures 3. The relationship with the assumption of constant returns is obvious from the above formulations. It assumes that every productive operation can be split up and realised on a reduced scale without changing the proportions of inputs and outputs. In practice..~ let k be the integral part -of f3. -We 'also speak of non-decreasing returns to scale when f(yl) :s. taking the vectors yl. 2y l. f3y/) > f3Yl = f3g(yz. by definition...Ya To characterise the second of the above assumptions. we shall consider that returns to scale are constant precisely when additivity and divisibility can be considered to hold. ·.54 The producer Assumptions about production sets 55 DIVISIBILITY. 0 (or y 1 E Y) and a > 1 imply f( ay l) :s. then the assumption of constant returns to scale is clearly satisfied.I)yl successively for yZ and thus proving that 2y\ 3y\ . . CONSTANT RETURNS TO SCALE. or non-increasing returns to scaie~ with the ~-ti(. since f3y is feasible whenever Y is feasible. The production set bounded by r I relates to constant returns to scale. y/). If the vector yl defines a feasible production (yl E Y or (fyl) :s. constant returns to scale imply that g(f3Yz. in Figure 6. On the other hand. additivity is not necessary. although rigorously. . 0)..E.k)yl is feasible. we often speak of 'non-increasing returns to scale' rather than of divisibility.. ..~e. .._-_.. 0). then the vector f3y l also defines a feasible production (therefore f3y l E Yandf(f3y l) :s. there is convexity if the set Y contains every segment joining two of its points.. . If the function 9 is homogeneous of the first degree.. and the other two sets are.L'0:'bjch _Vj_~_ shilJL shQ!~ be concerned. f3y/) = f3g(yz. it can be said to be rarely satisfied. then the vector ayl also defines a feasible production (therefore ayl E Y and f(ayl) :s.a)yZ defines a feasible production.Obviously the constant ~urns defined by this assumption imply divisibility. This assumption specifies that the volume of output can be changed without changing the return with respect to any of the inputs.. the return with respect to the input I in the productive transformation y' is. Con~~ditivity and divisibility imply constant returns to scale. repeatedly.~~i1. Obviously divisibility and additivity imply convexity. This assumption is much less generally satisfied than the previous one. the set bounded by r 3 is not convex. 0) and if f3 is a. bI = YI o Fig. 6 > Yl = g(f3Y2' . we (. Similarly. Since the null vector naturally belongs to Y. But this indivisibility may vary in its degree of effectiveness and in many industrial operations it appears negligible.. For every productive operation there is certainly a level below which it cannot be carried out in unaltered conditions.. .

To represent this situation.I I g~k dy" dYk (22) "=2 "=2 k=2 and (23) Here gj...)' negative definite or negative semid. convexity implies that the IJIatrix G" of the second derivatives ghk i... O. while additivity implies constant returns. is a general form of the assumption of non-increasing marginqlrc!urns. Equilibrium for the firm in perfect competition When dealing with the consumer. Equilibrium for the firm in perfect competition 57 this. In some branches of production. In this situation we usually find that production is carried on by a relatively large number of technical units functioning in similar conditions.. also called the marginal productivity. notably in Chapter 7 when we shall consider economies involving a large number of agents.).= PhI'" = I .. we shall deal with the case of a function of the form YI y? = g(Y2.I + Cl. This distinction will be taken again later. divisibility can be considered to be approximately satisfied to a sufficient extent. Allais uses the term 'differentiated sector' to cover all productive activity of this .= Ph b" - I . . . . Y? + CI._-~_. In other fields.!? scale and convexity. Expanding the right hand sides of (20) and (21) up to the second order.Yh)' We should point out that diminishing marginal returns and constant returns to scale are not contradictory._--_. Y?) and Y? '* + dYI = g(y~ + dY2' . taking the null vector for y 2 ... we reduced the problem of choosing the best consumption complex to that of maximising a utility function..ret~rns. _. we obtain 1 (1 + e) 1 I dYI = I 0.e) IS certainly negative if the dy" are sufficiently small).2 ' . . this fact may explain why we choose .g~) :::. The presence of considerable indivisibilities may explain the appearance of production functions with increasing returns to scale for which the assumption of non-increasing marginal returns is not satisfied. Similarly gi:k is the value at yO of the second derivative of g with respect to y" and J'k. it can be shown that. dyz. PY = I . 4. < I. Subtracting (22) multiplied by CI.1] . On the other hand. We shall now assume that the firm tries to maximise the net value of its production: I I I I I h = 2. Allais suggests that we distinguish two situations. . Y3. then the assumption of convexity holds. M. (25) . then yO + (X dy i~ a possible vector..) Convexity has consequences for the second derivatives of the production function. YI.!r!nd_~~visibility imply both constant~turns. Let us assume that the second derivatives of g are continuous. 0 (24) (the multiplier CI. Conversely.-- = g(y~.(CI. . considerable indivisibilities exist.kind. In particular. To investigate these consequences. and taking account of the fact that 0 < CI. YI)' (5) Consider two infinitely close vectors yO and yO + dy which satisfy (5): (19) 0(. . . Allais assumes that a single firm exists in each such field.Yh) The marginal return to h(ogjoa h = . The fact that certain factors available in limited quantities have not been taken into account explicitly in the formulation of the model obviously does not affect the marginal returns to the other factors._.!fini~~:------·"-·'·····--·"-·----'---'''-'''---------------..gj. The technology of this branch satisfies the assumption of constant returns to scale. . if G" is negative definite for any system of values given to Y2..functions for which returns to scale are diminishing._---~-~-_..= I Ph a".. Y? + dYI)' (20) If 0 < < I. we have 1 1 I I g~k dYh dYk h=2 k=2 :::. is therefore a decreasing function of the quantity of input h used (ah = . we need only apply the property of convexity. The market for each of the goods produced is then served by a very small number of very large technical units. dYI)' (21) CI.. . M. and taking account of (19). .l\ + . from (23). we return to the two reasons mentioned earlier for departures from additivity and divisibility. it therefore satisfies Y? + (X dYl :::. The two numbers e and 1] are infinitely small with the dy". is the value at yO of the first derivative of g with respect to y". I.._. o( . g(y~ + CI. d. this condition implies To conclude our discussion.Also. . as can be verified from the function YI = -JY2 Y3. M. The condition on G". all of which constitute what he calls the 'undifferentiated sector'. therefore non-increasing margina~._--~~<~-------.._~~_~_i~i~i. which we have just established..56 The producer that is. Since a priori the dy" can have any values.

we can easily determine equilibrium for the firm. BasicaIly. (26) (In what follows. Of course. . In particular.the price of each good is perfectly defined and exogenous for the firm. we shaIl later discuss the definition of 'profit' when time and uncertainty are taken into account. We shall not do thIS. or dispose of any quantity it has produced. it is assumed here that the fh are not all simultaneously zero. They would on the contrary be must. . Jot/mal oj Economic Literature. YI). we should no. at this price. see H. the microeconomic theory with which we are concerned considers the behaviour of the firm to be motivated by its desire to realise the greatest possible profit subject to the constraints imposed by technology and the institutional environment. It foIlows from the remark at the end of Section 1 that the ff< are not negative and consequently that A is positive. but they lead to an idealisation that may look strong with respect to an essentially complex reality. )'J) = O.w investigate the existence and uniqueness of equilibrium. inadequate for building a 'theory of the firm' that could serve as a gene. Dreze.and if.1 (27) where f(. (29) conditions (27) become PI = A t r and Ph = . we shall have explicitly introduced time and uncertainties. for this purpose. this is an abstract model of real situations. mention here the existence of a general equilibrium theory for economies with labour managed firms. On this subject see 1. which in any case raises some difficulties of principle (see the footnote at the start of Section 6).1'2' . Moreover.1'I' . . At the end of this chapter we shall discuss the case of the firm in a monopolistic situation and in Chapter 6 we shall briefly consider the formulations proposed for other situations of imperfect competition. if the production function is f~ f. Econometrica. t Concerning the difficulties faced by the theory of management of firms and the man~ references dealing with it. We need only maXImIse py subject to the constraint f(.58 The producer Equilibrium for the firm in perfect competition 59 This expression.t Adopting the assumptions of profit maximisation and perfect competitIOn. and using a production function representing the technical constraints. and therefore independent of its production decisions. ObvIOusly we also assume that the price vector is not identically zero. . 'Some Theory of Labor Management and Participation'. Again. equalities satisfied in the equilibrium. I repeat that these hypotheses are introduced here in order to permit the building of a general equilibrium theory and that. When in Chapters 10 and 11. This model is clearly inappropriate to the 'undifferentiated sector'. so th~t the firm.2. Thus.. it assumes that the demands and supplies emanating from other agents are completely flexible so that they can react instantly to any supply or any demand emanating from the particular firm. In fact. Conditions (27) imply Ps (28) Pr In the equilibrium. However. For our present purposes it is sufficient that the assumption of profit maximisation seems to afford the best way for a simple systematisation of the behaviour of firms. The objective of the fi. Maximisation of (25) subject to the constraint (26) IS a sImple case ot the classical problem of constrained maximisation. no alternative has so far been suggested which stands up to examination and can provide the basis for a general theory. . the hypotheses of profit maximisation and perfect competition have the advantage of being simple. The necessary first order conditions for a vector yO to be a solution imply the existence of a Lagrange multiplier A such that Ph=Aj~ h=I. we consider the firm to be in a situation of perfect competition if: . So we shall go straight on to consider the margmal . June 1979.. We must remember that the microeconomic representation considered here aims at a theory of prices and resources aIlocation not at a theory of the management of the fir~~. 'The Mlssmg Lmk: Micro-MIcro Theory. the marginal rate of substitution between the. .. we assume that no price p" is negative.t Also.) If we follow the same approach as for consumption theory. so that its actions have no influence on prices. the firm can acquire any quantity it requires of a good. they may provide an admissible first approximation. Leibenstein. YI = g(Y2' Y3' .ral conceptual framework for the discussion of the many problems concernmg decisions to be taken by business managers. has been subject to criticism. we shall also understand that strictly speaking perfect competition implies a much richer market system than the one actually prevailing. loses nothing by limiting itself to technically efficient net productIOns. ~wo com-II modities rand s must equal the ratio of the prices of these commodltles. some criticisms arise from misunderstanding of the wide generality of the model under study. is the value at yO of the derivative of f with respect to Yh' For the application of theorem VI of the Appendix. In order to avoid the same errors...nn is then said to be maximisation of value added per worker rather than maximisation of profit. November 1976. This assumption.. however. it assumes that the firm is small relative to the market. which is the amount by which the value of outputs exceeds the value of inputs also defines the 'profit' that the firm derives from production. 1. adopted in all theories of general equilibrium.Ag~ for h f:.

(35) where fh is the value at yO of the derivative offwith respect to Yh' In view of (27).inputs are increased by the quantities y~ do:. The profit associated with y1 cannot exceed the profit associated with yO.. . But its proof is much shorter. The case of additional constraints We have seen that the production function may be insufficient for complete representation of technical constraints.yO) dt] ~ O. the second order conditions imply more simply that lI.!:!iE1:>!il. . The maintenance of equilibrium for this sector demands forms of institutional organisation other than perfect L. With the general form of the production function.TIieT6110Wiii"gpropertYlliemore ma-tcnesproposition 2 "ffi<::I1apter:2.. In the particular case of the production function (29)..60 and so The producer The case of additional constraints 61 h=2. (33) then follows from (~1).. 3. Thus competitive equilibrium is incompatible with such increasing returns to scale. . these conditions require _g~=Ph competition (see. Consider a vector y 1 that is technically possible. we see that the multiplier of do: in the expression for dyddo: is / If dt tends to zero.n ofnon-increasing marginal returns.d derivative of f with respect to YII and Yk (see theorem VIII in the Appendix).ll!1l!t~-PQiI!!~(I!~ PIQcLuction st:Lwhere reLl. If we consider a limited expansion of dY1 and ignore the case where the second order term is zero. hence: I= / g~k dYIl dYk ~ 0 2 f[yO (33) ·~~~-'--~-. As in consumption theory.. y/) = ° (36) . Let dY1 be the corresponding increase in output. 0 - I1 for every set of dYh's (where h = 2. which is the required result.y~)f~ ~ 0.. I h. But f(yO) = 0. (26) say.. we can find the necessary. We can also now consider the inverse problem and prov~hat t"~_n:!~iI!<l. (32) where. we shall discuss briefly the treatment of cases where additional constraints must be added.! conditions (27) are sufficient for an e"quilil:lrTum-or"the-ftrD.0 0 gllkYhYk' </J(Yl> yz .'k denotes the value at yO of the secon.3.. then yO is an equilibrium for the firm. . (30) P1 The marginal productivity of commodity h must equal the ratio between its price and that of the output. Suppose first that the constraints are represented by the production function (26) and a second condition: " h. Section 6). . of course. the case of monopoly in Section 9 below. and consequently L h=l I (y~ . .k ~?POSITION lJ If the technical constraints are represented by a differentiable production function defining a convex set Y and if the vector yO satisfies (26) and (27) with an appropriate positive number l. for example. but apart from that may be any vector: (34) f(y 1 ) ~ O.k=Z It cannot be positi vewithout contradicting the necessary second order condition. whzch zs therefore sati~jied at an equilibrium for the firm. We can say that the returns to scale are locally increasing if dyddo: is an increasing function of do:. . 5. this inequality holds in the limit. I). J. we can always associate with these dYh's a number dY1 such that (32) is satisfied.l.k= 11=1 / f~k dYh dYk ~ 0 1 (31) for every set of dYIl such that I / f~ dYh = 0. Without going into details. So we come back to the assumptio. second order conditions for a profit maximum. For.j:-t~--=~-'-- + (i - yO) dt] ..y~) ~ 0. or the management rule for certain public services given in Chapter 6. which are often characteristic of the sector in which very large production units predominate. ifilie assumption "aTeonvexlty is·satiSfiecl. Th~econd order conditions reveal aE important point: the firm cannot be~~l'~itive_~. .-relating" to the consumer. . and since l is positive. Because Y is convex.. (35) implies L 11=1 I Ph(Y~ .dt)yO Let dt be a small positive number.f(yO) s.![!}§" the !o scale are locally increasing. and so f[yO + (y1 . Y? do:. the vector (1 + dty 1 is technically possible.. Let us take the case of the production function (29) and assumetha:tlrom~yO.

(37) I I / which replaces (27). fl. the multipliers fl. /(2 = 0 and fl. . Suppose. System (41) reduces to system (27) exactly as if the constraints (40) did not exist. /(1 = 0 and flz ~ O.l . the multiplier A must be positive and the equilibrium yO must strictly satisfy f(yO) = O.2 for Ph = Ai~ (41) P3 = Ai. fl.ayg (the point M on Figure 7). the marginal productivity .2. If PI or P2 is positive.2. + aj~) (39) This new system has the same form as (27) provided that goods 3 and 4 are replaced by a composite good one unit of which consists of one unit of good 3 and a times one unit of good 4.l ~ O./3 + af4 is then the partial derivative off with respect to the composite good.aY3' (40) (Goods 3 and 4 are inputs. = 1. but it is usually inappropriate for the discussion of general equilibrium.l. t Similarly. The function to be maximised is P1Yl + pzYz + P3Y3 the constraints are + P4Y4. . 7 B Eliminating fl. Y3. system (27) is replaced by Ph = Ai~ { P4 ~ Aj~ h o~ - Y4 ~ .aY3 ~ O. h = 1. + and P4 aj~) ~ Aj~ . System (27) becomes Ph = Aj~ P3 + ap4 { P3 ~ Aj.3. for example.fl.l and fl2 must be non-negative.= 1. + afl.2 where each of the multipliers A. After elimination of fl.2.fey!> Yz. Let us return to the example of four goods and the additional constraint I I I I I I Y4 = aY3.2..y~ < .62 The producer The case oj additional constraints 63 After introduction of a second Lagrange multiplier.) Here we have a case for application of theorem XI of the Appendix.. System (37) becomes Ph = Aj. = AU.2. since goods 3 and 4 may be produced by two distinct firms. if the production function takes the form (5). and must be zero when the corresponding constraint is a strict inequality. (38) I I I I I I I I which expresses strict proportionality between two inputs.. that there are again four goods and. and the proportion of 4 with respect to 3 is bounded above. (43) = AU. { P4 = Aj~ + for fla fl h = 1.2 be the corresponding Kuhn-Tucker multipliers. we obtain Ph = Aj~ { P3 + ap4 h = 1. Y4) ~ 0 . apart from the production function. the two constraints Let A. no insurmountable problem arises if we take account of constraints expressed by inequalities. P3 = Aj. I. The necessary conditions for a maximum are h . We can then distinguish three cases: (i) If the equilibrium is such that 0 < .Up~. or consumed by other agents in a proportion other than <x.94 of good 4 is less than or at most equal to the price ratio P4/Pl' (iii) If the equilibrium is such that y~ = ay~ < 0 (point A in Figure 7).l and fl2 are zero. t The introduction of such a composite good raises no difficulty when we are considering the firm in isolation.Y4 ~ 0 { Y4 . see Figure 7.l and fl. h = 1. (42) ---- . Does such a substitution have much effect on our results? Not necessarily. the first order conditions become: Ph = Aj~ + !. A relatively simple alteration in the properties is sufficient in some cases.2 { P4 = Aj~ + fl. (ii) If the equilibri um is such that y~ = 0 and yg < 0 (point B on Figure 7). In particular. I I I o Fig. as we shall assume.

'~t involve p and the function to be maximised is ~o~ogeneous II1 p. We shall write this law rth(P1. it also maXimIses rJ. but we can now identify the irtdividual marginal productivities of inputs 3 and 4 with respect to output l. then every proportional vector exyO will also be a solution (ex > 0).2. by definition. P I h t' t E Y orf(y) 0 hIS is an 0 vious property since t e cons ram. these equations will not generally have a solution. if the production function satisfies the assumption of constant returns to scale and is expressed in the form (5) or (29). . Now. Similarly. functions can be defined only for a subset of the values that are a priori pos~ Tor-. profit is proportional to the volume of production.64 The producer Supply and demand laws for the firm 65 This brings us back to (39). k = 1.. In the context of the perfect competition model.I1IS 1e~. If the Ph are chosen freely.-l/Y" (ii) the determination of the volume of production. we can introduce a composite good for the interpretation of the last equality. necessary for equilibrium and also sufficient in the case of convexity.. this assumption is more restrictive than appears at first sight.So the term'supplyC~pply functions' is used. In short. 6.to and for any multiplication of these prices by the same pOSItIve PuPz. .. (44) To establish this property. :wo ' ts h al1d k are complements if thiS denvatlve IS 'pOSItIve.. d This property shows that we can say unambiguously whether two goo s are substitutes or complements for the particular firm . Agam It can be described as 'the absence of money IllusIOn. are homogeneous of degree zero and can therefore be expressed' as functions of the 1. to incrc:ase the input of factor 3 without changing the input of factor 4 is possible but not worth while. only zero production gives an equilibrium which does not obey the marginal equalities (30). F" [ [f']' 0 I'J_1 . If yO maximises py subject to the' constraint. IS clearly symmetnc. being fixed. . PI). . PI' the set Y of feasible productions.-I!Y. We n~ed only ~oOk at the sign of the partial derivative Ol]h/OPk' More pre~lsely: w~ sa~ tat..2 variables YZ/Y" . .!({.n. " Y.t a~ect eqUIhbnum. the matrix (47). which we assume here to eXIst. The decision to produce can be split into two stages: (i) the choice of the technical coefficients Yl /Y" . this homogeneity 0 . y ~ .. the two stages are independent of each other and. there are I . . yO say. ~ the net supply functions are differenti~ble: we can :haractense thiS 'substitution effect' of h for k by the pa:tial den~atlve of I] h With respect to Pk' Property (ii) then expresses the followmg equalIty: prMe~. the supply function for commodity h defines how the firm's output of this good varies as the prices of all goods vary. the derivatives g{. Now. once the best technical coefficients are chosen. t. whereas to increase the input of factor 4 without changing the input of factor 3 might be worth while but is impossible. then any level of production is optimal. If profit is zero.. .s no... " . we differentiate the system consisting of (27) and (26) and obtain A \ rt f~k dYk + f~ dA = dph h = 1. Y. Pz.t which can be written in matrix form: t In fact..and the hth colum. ~et. and f.. and is unique. assuming that yO exists. shows that the choice of numeraire doe.S~et su o The theory of the firm must lead to some general properties of supply and demand functions.1 equations (30). or the production function f. (45) I h= 1 f~dYh = 0. . In the case of constant returns to scale. is positive. If it is positive. as happened with the theory of the consumer. and are substitutes if it is negatIve.. The substitution eU!ct of h for k is equal to the sUbstltut~ fpr h Consider the increase in the supply of h when the pnce 0 Iml. outputs or t wo mpu . We see that the marginal productivity of input 3 is at most P3/P1> and that of input 4 is at least P4/Pl' In fact.-~~. AF" [ [fT 0 I'J [dYJ = [dPJ. .. f I f ctions Just as in consumption theory. supp Y ~n. . . Supply and demand laws for the firm I unction is homo eneous 0 de ree zero with res ec.t . no equilibrium exists since it is always advantageous to increase production. For example. In the particular case where the Ph are such that a solution exists. which. is equal to supply for an output and to demand with a change of sign for an input. namely fi/f. R'. The most modern versions of microeconomic theory take account of these difficulties: net SUR-pIx. dA 0 (46) (47) with the obvious notation. We shall deal with these two functions simultaneously by considering net supply. If it is negative... which proves the equality. . consideration of additional constraints entails some modification in the equilibrium conditions but makes no basic change in their nature. the demand function for commodity h defines how the firm's input of this commodity varies. .2. In economic terms these formal difficulties have the following significance. This equality shows that the left hand side of (44) is the element on the hth row and kth column of while the right hand side is the element on the kth ~ow . The net supply law for commodity h is therefore that law which defines Yh as a function of the P1> P2. for every vector P belonging to an I-dimensional domain of We can easily establish the following three Ol]h 0Pk Ol]k 0Ph h. l. .py when rJ.

pI and pZ say. . The theory of the firm is often built up on the initial basis of the cost function.. but is subject to criticism on two counts. we obtain (pi _ pZ)yZ ~ (pI _ pZ)yl or YI = g(yz. the net supply of this good cannot'll diminish.. Section 9). is not necessarily equal to Ph/Pl' It is equal to P. However.. we can write plyZ ~ plyl (48) and also ® pZyl ~ pZyz the cost function changes when these prices change. Consider two price vectors. The cost function relates to the quantity produced YI.. as we assume here.. the solution is obtained by determining values of A and of yz. . property 3 in Chapter 2. . Cost C is defined as C = h=Z L Phah I = - h=Z L I PhYh' (53) We need only replace the Yh in this expression by their values in the solution of (52) when we want to determine the cost function. This greatly simplifies the analysis. the inequality becomes: (p~ . PI = A + ~ { Ph = . Since yl maximises ply in the set of the feasible y's and since yZ is feasible. The reasoning is similar to that used for consumer demand (cf. In the first place. . Since our aim is to lead up to the study of general equilibrium. Y3. an examination of cost functions reveals certain useful classical properties which are simple to establish at this point and may be needed later. Yl)' (49) (29) :* (pI _ pZ)(yl _ yZ) ~ O. yl and yZ..' yz) = ~l { Ph ... Since we restrict ourselves to the case of only one output.. In the second place. so that When the firm minimises its cost of production.!PI if YI is the optimal production for the firm selling on a competitive market.YI' Here the system of first order conditions (37) becomes This is the general form of the relation of comparative statics.66 The producer Cost functions 67 When the price of a good increases. hfor example. we can take the production function as or equivalently. a production theory based on the analysis of costs is out of place in a general equilibrium theory which treats prices as endogenous and not determined a priori. the minimum value of the input mix which yields this production.. we must first find the combination of inputs which allows production of a given quantity Y1 of commodity 1 at minimum cost. But for freely chosen YI' in most cases it is not equal to this ratio. _ pZyz ~ _ pZyl.3. In particular. . If. which must be obeyed in the comparison of two different equilIbna for the same firm. the first order conditions are sufficient for cost minimisation. Cost functions Suppose that the prices Ph of the different commodities are given and that the firm produces only one good. . . The first equation allows us to find /1 and is of no further use.3. The production set or production function are more fundamental since they represent the technical constraints independently of the price system. the relationship between the value of input complex and the quantity produced depends on the prices Ph of the different inputs.Agh for h = 2. [. and two corresponding equilibria. the good I to fix ideas. (50) Before defining the cost function.yt) ~ O. . We assume here that the markets for inputs are competitive so that the Ph are given for the firm (h = 2. I). the marginal rates of substitution of inputs are equal to the ratios of their prices... (52) 7. We can also proceed directly on the basis of finite differences.. . YI which satisfy This establishes property (iii). Y3' . but the marginal productivity of an input..3. For the proof of this property we can use the second order condition for an equilibrium and establish that the partial derivative of/lh with respect to Ph is not negative. g(y:: Y3' . which relates the value of the minimum of C with the production level YI (the Ph being . which makes the result clearer and more general. we must start with production sets or functions.. so we must maximise profit subject to the constraint that Yl = YI' This is a particular case of the problem discussed at the start of Section 5 where ¢(y) = YI . Adding (48) and (49). if pI and pZ are identical except where price P" is concerned.pt)(y~ .2.Agh h . l.

equations (29) and (30).. t This function is often assumed to have the form of the curve C in Figure 8. are transformed into first order conditions for profit maximisation. (ii) Choose YI so as to maximise profit (pdl .t = - . Let us differentiate (53).. .k~ I I I. is concave downwards at the start: thIS corresponds ~ to the range of values of output for which indivisibilities are significant and marginal returns are increasing. . When marginal cost is equated to price PI' the first order conditions for cost minimisation. differentiating the first equation. The first order condition requires PI = C'(jil)' (54) C' measures the increase in cost resulting from a small increase in production. .68 The producer de Cost functions 69 (55) considered as given). and is therefore the 'marginal cost'.. we assumed that the quantities of all the factors could be freely fixed. This latter assumption is inappropriate to factors such as the work capacity of the managing director. The classical curve of the. z Since marginal cost A is positive. Equation (54) shows that.. A equals the marginal cost. The functIOn g is then homogeneous of the first degree.~z L I p" dy" or. C = AYl' This equation. taking account of the definition of C and the marginal equalities (52).~z t g~ dy" = dYl A ktz g~k dYk = 0 (56) h = 2.. . = 2. cost function. equations (52). So the case of constant marginal cost is not necessarily frequent in relation to a firm some of whose factors cannot vary. in competitive equilibrium.. I.. take account of Multiply the hth equation by dy. Y3. t The term 'cost function' is sometimes also used for the function that relates C to 11 and to Pl. sum for h the first equation: we obtain Fig. determine for each value of YI the Yz.. hence. (58) dYl which is the required result. . (See below the distinction between long-term and short-term costs. So a cost curve derived from a production function with non-increasing marginal returns is concave upwards. We can also verify that the assumption of non-increasing marginal returns impli~s tha~ marginal cost is increasing or constant. 8 When looking for the equilibrium of the firm. as exhibited in Figure 8.Yh = YI. the assumption of non-increasing marginal returns implies dA dA' dYl ~ 0 or ~ 0. we can work in two stages: (i) Define the cost function. I. that is. When defining marginal cost. that marginal cost is increasing or constant.) .3.~z L g~dy" = AdYl' This equation establishes that ). taking account of (52) and.. The second order condition requires that the second derivative of the profit is negative or zero. that is. \1 keepmg pnces constant: If ldAg~ + dA dYl r . together with (55) shows that A. Let us differentiate (52). and so + A h. PJ ... h~2 I I g.PI' t We saw that the assumption of constant returns to scale would usually not hold if all the factors of production were not accounted for in the model.3. We note also that marginal cost is rigorously constant when the prod~ction 1\ function satisfies the assumption of constant returns to scale. marginal cost is equal to price of the output.. Yl which minimise cost and find the value C corresponding to this minimum cost. in (52). keeping prices p" constant: dC (57) g~k dy" dYk = O. We shall verify that. . . in particular. the expression for cost.C(YI))' The solution of stage (ii) is obvious. equals marginal cost. which a priori is a function of Yb is in fact a constant (always assuming that the p" are fixed). = A ..

7 ... often distinguish between long-run decisions relating to the entIre orgamsatlOn of production (choice of equipment and manufacturing processes) and sh?rt-run decisions relating to the use of an already existing productive capacIty. so that the ambiguity disappears. Thus for the s. it is immediately obvious t'hat average cost is increasing or decreasing according as it is greater or less than marginal cost (a typical curve c appears in Figure 8).ingle gO?d. to fix ideas. It is sometimes convenient to give a diagram representing the last stage in profit maximisation. YI). c .price PI)' ." " " . If we differentiate c with respect to YI.. this is likely to arise only in two ways. In the first place.. P3. then the others also correspond to equilibria. that capital equipment is represented by a. YI_I> cost C* and marginal cost). In practice. Let the curves c and y represent respectively variations in average cost and marginal cost as a function of YI for given values of Pl. the first point cannot correspond to an equilibrium since it does not satisfy the second order condition. o Fig. we sometImes consid~r that some firms actually behave so as to provide an exogenously determined output and minimise their production cost.ue. which was limited to finding necessary conditions for a profit maximum at a point yO for which constraints other than the production function do not operate. As we said previously./ i 1 I " 1 em ------~-~-~-~---. as we assumed earlier when we said that y? corresponds to the equilibrium? Ambiguity may exist if several points on y have PI as ordinate. Yl of output :vhen Y. P3. System (52) then applies directly to the equilibrium for the firm. (Of course. together WIth the multiplicity of equilibria in (ii)..". S~ch situations can easily be analysed using the principles applied above. which mayor may not be defined uniquely.. The point or points with ordinate P I and lying on the non-decreasing part of y may not correspond to an equilibrium if it is to the interest of the firm to have zero output Y I ' This situation arises ifPI is less than the minimum average cost Cm and if Yl = 0 implies zero profit. if one of them corresponds to an equilibrium. .. .* obey the system . we. since the points considered then give negative profit. Pi. if the whole curve y lies below the ordinate corresponding to PI> there is no limit on the increase of profit and it is to the interest of the firm to go on increasing production indefinitely. As before. one on the decreasing part and the other on the increasing part of the marginal cost curve. in practice it would come up against a limit sooner or later. there may be two such points. but only some of its inputs. Short and long-run decisions Cost minimisation has just been presented as a stage in profit maximisa~ion. all the points on this flat section give the same profit. ""PI' the value of PI may be such that: (i) the firm should choose YI = 0 (low price PI). Suppose. Similarly.(I~I~.ly (hig~. (ii) the firm should choose a finite output Y?. In fact abandoning the strict model of perfect competition. Also. the others being predetermined. 9 ~ (iii) the firm should increase production ind~finite.) To sum up..~'-----~--. we see that inputs Y2' Y3' .. Let thiS functIon be C (y I. the inputs relating to capital equipment are fixed..ame firm. in some contexts.. So for short-run decisions.70 The producer Short and long-run decisions 71 In addition to total cost C and marginal cost C' we often consider average cost per unit of output. the firm does not choose all. the /th. at the ordinate PI the curve y may be flat (in particular.The equilibrium point yO is determined by the abscissa Y? of ----------------~' . Are these conditions also sufficient. . Let YI be the predetermined value of YI' The short-run deCISIOn conSIsts of profit maximisation subject ~o the constraint YI = YI' The s~ort-run cost function relates cost C to the ~al. Finally. we saw that marginal cost is constant if the productwn function satisfies the assumption of constant returns). are sufficiently real possIbIlItIes to ~ake us avoid trying to prove for producer equilibrium a general property ~~ ex~sten~e and uniqueness corresponding to that stated for consumer eqUIlIbnum 111 proposition I of Chapter 2. . ." the point on the curve }' whose ordinate is PI' The profit is then Y? times the difference in the ordinates of the points on y and c with abscissa Y? Examination of the figure rounds off the preceding analysis. namely c = C/h. s. for given values of Pl. the existence of situatIOns (I) and . 8. but the chosen cost function ignores this fact.I = YI' t~e _ot~er inputs Y being fixed so as to m1l1ImISe cost.

A*g~ 1-1 h h = 2.' TC n . is satisfied. . Monopoly The formal approach developed so far is more or less easily transposed to institutional situations that differ from perfect competition. p. long and short-run average costs are tangential. are independent of the decisions of the firm. At this point. . It knows that price Ph depends on the quantity a h = .. We have to adopt a formal model other than that of perfect competition. gives the value YI for J'l.. It can obviously be eliminated if prices Pz. long and short-run marginal costs are equal. since if existing equipment coincides with what the firm would choose in the long run in the same where n I is the function defining the price at which the monopolist can dispose of the volume of production YI' lt may also happen that a firm is the only one to use a factor h (for example. I pomt on yC whose ordinate is Pl' The long and short-run average cost curves generally have a common point corresponding to the value of YI for which the solution of (52). I I . . t We can represent this demand by a relation between PI and YI: (61 ) which replaces (55). for example.. we obtain dC* = A* dh . Classical monopoly theory represents this situation starting from the hypothesis that the same price Ph will apply to the exchange of all units of commodity h but that this price will depend on the quantity )'h that the seller will supply. then short and long-run equilibria must naturally coincide. In the applied study of market structures a firm is said to have a monopoly position on the market for commodity h if it supplies alone this commodity and if demand comes from many agents who are individually small and act independently of one another. . defining the I eC ~ ------~-----.. The firm facing such a situation necessarily takes account of the fact that the price at which it will dispose of its output depends on the quantity which it puts on the market. 10 long-run cost. The long-run equilibrium value ofproduction for price PI is determined as the abscissa yf of the point on yL whose ordinate is Pl' Also let cC and yC be the short-run average and marginal cost curves. . (59) C* = - L PhYh h=2 .(A*g. 3. I . Let yy be this particular value of YI' At yy. We can no longer analyse its behaviour on the assumption that it considers price as exogenous. . We could also verify that. Hence. Thus the monopoly faces a demand whose quantity varies with the price of his product but is otherwise independent of his decision. the short-run cost cannot be lower than the long-run cost smce the minimisation which defines the former is subject to one more constramt than that which defines the latter.'/' II . . Suppose. YI-I. A priori. / ). For. The short-run equilibrium is determined by the abscissa yC of the . is made here for the sake of simplicity. The short-run marginal cost is again equal to the equilibrium value of the Lagrange multiplier A*. the equality PI = . 9.Yh t The assumption of independence of demand with respect to prices Pz.. . eL /1 I I I I I I I o I i yO yC yL . leaving for Chapters 6 and 8 the analysis of other situations.PIYI' Differentiating the first and last equations for given Ph and taking account of the intermediate equalities.A*g. price situation. when it is the only employer of labour in a town). lt is said to be in a situation of 'monopsony'. so that dC* = A* dYI = dC. this may seem an obvious result. Fig.72 The producer Monopoly 73 I I g(Yz. . p. We may briefly examine here the classical theory of monopoly. that is. the solution of (52) then satisfies (59) with c* = C. .---t~1 . the long-run average cost curve is the envelope of short-run average cost curves (obviously the same property holds for total cost curves).. if the markets for all goods except the first are competitive. YI) = Ph = . In any case. '. + PI) dYI.. . that the firm produces good I and sells it on a market where there are many buyers whose demand depends on price PI and not on other prices. I - 1.. . to determine the value of YI which maximises profit subject to the constraint YI = YI' we must add to (59) the condition that the marginal cost A* equals Pl' Let us illustrate this theory by a diagram in which the different cost functions are represented as a function of YI' Let cL and yL be the long-run average and marginal cost curves. I y.

We can easily compare monopoly equilibrium with equilibrium for the firm in perfect competition. in the competitive situation. the function 1tl + Yl1ti.. As in the case of perfect competition. is a Lagrange multiplier.. where R(Yl) denotes the firm's receipts from output Y1' Profit maximisation implies that Y1 is so chosen that (67) = y". .) = Af~ h = 1. In ?rder to investigate (64).. the firm in a position of monopsony usually employs a smaller quantity of the factor k than it would employ in competition. the good 1 being the firm's output... The marginal productivity of the factor h is no longer equal to the ratio of prices but to this ratio multiplied by a term depending on the elasticities relating to the factor h and to output. we shall consider the case where the production functIOn takes the form Y1 = g(YZ'Y3' ""Yl). and the curve (j representing marginal revenue.the case of perfect competition. as the inverse of the elasticity of demand (or supply) whichhoccurs in the mar~et for. .. is h= [ I I provided that 81 i= .. or net value of production..C(Y1)' We can write this expression in its usual form R(Y1) . which is thus reinforced... (64) taking account of the fact that Ph is the value of the function n and defining 81. I. Conditions (64) reduce to the first order condItIOns (27) obtained earlier. that is.2.. Y1 . We.g~ h = 2. whose price is greater than that actually asked by suppliers. the firm's demand 1] k can only decrease. However we shall adopt a rather different approach for an alternative presentation of the analysis. I). where n h is the derivative of nil and }. .y" of h.. For a pure monopoly. cost minimisation is carried out in exactly the same way as for a perfectly competitive firm and the cost ft:nction is exactly the same.) . = n~ d log nh d log /y. Equations (66) then reduce to the perfect competition equations except for the kth. the firm will fix its decisions as a function of a supply law p" = n. (63) Maximisation of. that is.(y.) (62) representing th~ behaviour of the agents supplying the factor h and indicating the pnce Ph whIch the firm must pay to acquire a quantity . I Pl(1 + 8 1 ) (29) R"(Y1) ~ C"(Y1)' (69) (66) Equation (68) generalises condition (54) obtained for the case of perfect competition. we shall consider the case where prices are non-zero and shall WrIte the above conditions in the form p. . we can maximise profit by means of a two-stage procedure involving first cost minimisation and determination of the cost function. note that the case of perfect competition corresponds to the particular sItuatIOn where n [ and nh are constant functions. For this reason it may be said to be in the interest of the monopsonist to adopt a 'Malthusian policy'.2. thi~ expression subject to the constraint expressed by the productIOn functIOn Implies the following first order conditions: h = 1.gk must equal pdp1 multiplied by the term (1 + 8k) which is usually greater than 1.74 The producer Monopoly 75 which it uses as input. We could apply the same reasoning to the case of pure monopoly where all the 8 h except 8 1 are zero.. . y". Since. 2. .(1 + 8. So we can confine ourselves to the second stage.1 in the equilibrium. For w. the good h because of agents other than the particular firm under consIderatIOn: 8" 1t1(Y1) . where .hat follows.. If it takes no account of the possible interdependence of Ph and the pnces of other goods. . .(1 + 8h) . I.. which we assume for simplicity.(y. as will .. As a function of y the profit. Suppose that 1ti is negative. the 8 h are zero. market demand and supply are perfectly elastIC from the standpoint ?~ the firm. Therefore we can deal sImultaneously with monopoly and with monopsonies concerning one or more factors by treating the case where the firm tries to maximise its profit and takes account of functions n h relating the price of each good h to its net production Yh (h = I. Figure 11 shows the average cost and marginal cost curves c and y. Consider first the case of a monopsony for which all the 8" are zero except that relating to a particular input k. Equations (64) imply p. and find the value of Y1 which maximises n.C(Y1). average revenue. The equilibrium is therefore the same as in a situation of perfect competition involving the same prices for all the goods except k..J (65) R'(Y1) = C'(Y1) and (68) In.. as well as the curve d representing the demand function 1t 1(Y1).

1 is a fixed number." . This is just the situation for a monopoly. In most cases. But we should point out that this condition may also be satisfied in situations where marginal cost is decreasing. which is possible in the "undifferentiated sector".. and if they are unable to band together in opposition to the monopolist.. . two remarks may usefully be made already at this stage... large and they are all of the same relative importance. the case of a firm that has a monopoly on each of the two or more independent markets in which its output can be sold.. However. We can consider R" as negative in the interpretation of (69) defining the second order condition for a maximum... since those who demand the proc.. _- "I . situations other than those we have considered can be dealt with by constrained maximum techniques. this result is similar to that encountered earlier for monopsony.. that is. it is conceivable that a ... : 1 I "b ~ ~.T I . monopoly equilibrium is determined by the abscissa y! of the point of intersection of y and (j. For example. I I ' . On some ... I I / I I I pO 1 1 I C . I I I o y. since then [.t' p. if it takes no account of the reaction of price PI to its supply YI.... They have no other possible attitude if their number is. second order conditions implying concavity of the isoquants in the neighbourhood of the equilibrium must be satisfied.. In particular it will be negative if there is constant elasticity of demand. Figure 12 shows an example for a firm with continually decreasing marginal cost. (j then lies below d. In the second place. We shall not pursue this line for the moment. If the firm behaves as in perfect competition.1) and R" to n'l(1 + [. but shall take it up again in Chapters 6 and 7.. It follows from the fact that d is decreasing and from the ~espective positions of d and (j that y! is necessarily smaller than y? The firm produces less in a position of monopoly than in a situation of perfect competition involving the same prices for it. Generally we can say that constrained maximisation is appropriate to the extent that all agents except at most one adopt a passive attitude. particularly wealthy consumer may have such influence on a market that he has a position of near-monopsony. 12 I Fig... '.76 The producer Monopoly 77 necessarily be the case except perhaps for an inferior good. R' is equal to n1(1 + [. Of course.. the marginal cost must be nondecreasing for y? to correspond to a true competitive equilibrium. When this is not so... 1 yO 1 o Fig. no equilibrium exists as long as the markets for the factors are competitive: but a monopsony for the firm may allow equilibrium to be realised... be realised which is not possible in perfect competition... profit maximisation leads to price differentiation.. t t We should also note that.1)' The second order condition is therefore satisfied for any situation where marginal cost is increasing.iuct accept as given the price which results from the firm's decision on production... According to (68). for example. More generally. _ I.__. In the first place. for the definition of the cost function. The study of monopoly has taken us outside the field of perfect competition. the firm releasing to each market a quantity of its product such that marginal revenue from each market equals its marginal cost over all its output.. . But imperfect competition is not limited to such situations. 11 At the point of intersection of y and d. the theory of imperfect competition cannot depend entirely on the constrained maximum techniques which we have used up till now. " .---.. . the equilibrium point is determined by the abscissa y? of the point of intersection of y and d. monopoly may sometimes allow an equilibrium to.! i c . taking the decisions of other agents as given. . it is clear that situations of imperfect competition may involve consumers as well as firms...

78 markets there are relatively few buyers and sellers. on others. Clearly it may appear very ambitious to attempt to deal with this. Other methods of analysis are necessary to deal with such cases. By providing an initial formalisation and by rigorously establishing conditions for the validity of classical propositions. and then to go on to the normative standpoint of the search for the optimum. Definition of optimal states Before fixing a principle of choice. 1. Optimum theory involves a rather simpler and more general model than the model on which competitive equilibrium theory is based. The classical approach would be first to discuss competitive equilibrium. We therefore change our perspective and attack the problems raised by the organisation of the simultaneous actions of all agents. keeping to the positive standpoint of the previous lectures. in the problem of the best possible choice of production and consumption in a given society. we shall reverse the order of these two questions. We are interested. optimum theory provides the logical foundation for a whole branch of economics. therefore. Preoccupation with the optimum underlies many propositions briefly stated by economists. But it is one of the ultimate objectives of economic science. 4 Optimum theory Up till now we have been considering the behaviour of a single agent. With the theory of the optimum we approach the study of a whole society. For our present investigation. However. We shall return to imperfect competition in Chapters 6 and 7 in order to clarify problems of general economic equilibrium. We shall then see how it relates to the theory of games. a 'state of the economy' consists of m consumption vectors Xi and n net production vectors Yj' . We must first find out what is meant by the 'best choice' for the ~ociety and go on to study the characteristics of situations resulting from this choice. coalitions take place. It seems plausible that the relationship of the two theories will be more clearly understood if those assumptions which are not involved in optimum theory are introduced in the later discussion of competitive equilibrium. we must again define what are 'feasible' states.

Master your semester with Scribd & The New York Times

Special offer for students: Only $4.99/month.

Master your semester with Scribd & The New York Times

Cancel anytime.