Econ Polit

DOI 10.1007/s40888-017-0082-2

Viability, reproducibility and returns in production

price systems

Enrico Bellino1

Received: 3 March 2017 / Accepted: 23 October 2017

 Springer International Publishing AG 2017

Abstract Production prices have often been conceived as those exchange rates
among commodities which allow their reproduction in the same quantities (or along
a proportional growth path). A deeper investigation of Sraffa’s price equations
reveals that this characterization is not well grounded. The notion of ‘viability’ of a
system is thus re-defined here in such a way as to allow the determination of
production prices for economies that are not in a self-replacing state (i.e. displaying
positive net products in some industries and negative net products in others). Via-
bility is here connected to the possibility for each industry to reintegrate the value of
the means of production and to obtain a uniform non-negative rate of profit. It
appears, thus, as a notion that mainly impacts the value side. This specification is
relevant for an approach where the quantities are determined separately from the
relations between value and income distribution. The focus on non-self-replacing
systems opens the question on how to handle this and all cases of systematic
changes in output levels without losing the property of persistence of production
prices, which permits them to be regarded as centres of gravitation. This issue is
explicitly linked to possible specific assumptions that can be made on returns. The
various views on this point are discussed here in relation to the main features of the
modern classical approach.

Keywords Self-replacing systems  Prices of production  Viability  Returns 

Modern classical approach

JEL Classification B21  B51  B54  D24  E11

& Enrico Bellino

enrico.bellino@unicatt.it;
http://docenti.unicatt.it/ita/Enrico_Bellino
1
Università Cattolica del Sacro Cuore, Via Emilia Parmense, 84, 29122 Piacenza, Italy

123
 Econ Polit

1 Introduction

The notion of prices of production and their interrelation with distributive variables
has been analysed in detail by Sraffa (1960) within a self-replacing system, that is, a
system which produces an amount of each commodity sufficient to repeat all
productive processes at least on an unchanged scale. Production prices have thus
been often considered as those prices that make it possible to reproduce all
commodities at the level of the entire system.1 But, if we only simplify the price
equations of a self-replacing system, we see that the reproducibility conditions break
down, even though the system remains able to determine the same set of semi-
positive prices which allow each industry to reintegrate the value of the means of
production employed and to obtain a uniform rate of profit among them (which can
be positive or zero). Consequently, a logical difference between the notion of
production prices and that of self-reproducibility can be established. To this purpose,
we will define a system as viable if it is possible to determine a set of production
prices which puts each industry in the condition to reintegrate the value of the means
of production employed and obtain a uniform (non-negative) rate of profit.
Alternatively, we will define a system as self-replacing if the quantities produced
by each industry are sufficient to repeat the production processes at least on an
unchanged scale. It will be easy to verify that self-replacing systems are viable but
not necessarily vice versa. The notion of viability comes thus to be characterized in
the sphere of values; this view is more suitable for Sraffa’s approach, where
quantities are determined separately from the relations involving commodity prices
and distributive variables. In addition, it allows us to see that production prices can be
computed also for systems which are not in self-replacing conditions.
The possibility that a viable system is not in a self-replacing condition was
merely hinted at by Sraffa in a footnote in the first chapter of Production of
Commodities by Means of Commodities. Moreover, in his ‘reply’ to Roy Harrods’
review of the book (see Harrod 1961), Sraffa (1962) disproves a conclusion drawn
by Harrod by modifying the numerical example provided in the first paragraph of
the book, in such a way to make it a system with a negative net product for one
commodity and a surplus for the other one (i.e. a non-self-replacing system). Far
from being an exception, the case of non-self-replacing systems represents a typical
situation of actual economies. Growing industries coexist alongside declining
industries as a result of many factors: changes in the composition of final demand,
technical progress, etc. These factors have been extensively analysed in the
economic literature,2 but further analysis is still needed to reconnect these situations

1
Piccioni (2000) and Ravagnani (2001) argue in detail against the foundations of this association. In
Ravagnani’s paper one may find a set of references where this view is advanced. For example, regarding
the system presented by Sraffa in his 1st Chapter, Schefold writes:
This model serves the purpose of, and derives its value from, clarifying in general the function of
relative prices in Sraffa’s system. Relative prices … represent the exchange ratios between
physical goods that make reproduction within a technical (methods of production) and social
(distribution) framework possible (Schefold 1989, p. 285, emphasis in the original).
2
The most articulate analytical framework developed to include these phenomena within the classical
theory of production is the model of structural dynamics worked out by Luigi Pasinetti (1981, 1993).

123
Econ Polit

with Sraffa’s price system. This is one purpose of the present paper, and will be
developed in Sects. 2–4. In his book Sraffa carried out his analysis in two steps: he
considered subsistence systems in Chapter I and surplus systems from Chapter II
onward. These steps cannot be replicated in the same way here, because in this
paper we distinguish between viable and self-replacing systems. A distinction will
thus be made between systems with a null maximum rate of profit, R = 0, described
in Sect. 2, and systems with a positive maximum rate of profit, R [ 0, described in
Sect. 3. In Sect. 4 the cases of non-basic products and of joint production are briefly
addressed.
The analysis of non-self-replacing systems raises immediately a crucial question,
which is common to all situations where output levels change systematically: how
do we guarantee the persistency of production prices, a property that must be
preserved in order to regarded them as centres of gravitation? This is quite a subtle
issue on which the specialized literature is not unanimous, since it concerns the
opportunity of introducing assumptions concerning returns. After a brief description
of the various views, the assumption of constancy of technical coefficients with
respect to changes in output levels will be appraised in light of Sraffa’s overall
analysis. It will be here argued that this specification can be done without prejudice
to the main purposes pursued by Sraffa. This is the second goal of the present paper,
and will be dealt with in Sect. 5. Section 6 provides the concluding remarks.

2 Self-replacing systems vs. viable systems: the case with R 5 0

Consider a self-replacing system similar to the one presented by Sraffa at the

beginning of his book (Sraffa 1960, p. 3):3

Example 1: a self-replacing system

6 qr. wheat  4 t. iron ? 8 qr. wheat
2 qr. wheat  9 t. iron ? 13 t. iron
= 8 qr. wheat = 13 t. iron

The corresponding price equations are

6pw þ 4pi ¼ 8pw ; ð1aÞ

2pw þ 9pi ¼ 13pi ; ð1bÞ

whose solutions are pw = 2pi. Clearly, if we simplify Eq. (1a), the system
3pw þ 2pi ¼ 4pw ; ð2aÞ

3
Sraffa’s original example has not been considered here for reasons that will be clarified later.

123
 Econ Polit

2pw þ 9pi ¼ 13pi ; ð2bÞ

has the same solutions. System (2a, 2b) may be imagined to be the price system
‘generated’ by the following pair of production processes:

Example 2: a non-self-replacing system

3 qr. wheat  2 t. iron ? 4 qr. wheat
2 qr. wheat  9 t. iron ? 13 t. iron
= 5 qr. wheat = 11 t. iron

This is not a self-replacing system: the wheat industry displays a deficit while the
iron industry displays a surplus. These examples show that the conditions which
guarantee the existence of an economically meaningful (i.e. positive) solution of the
price system and the self-replacing conditions do not always coincide.
For our purposes, it is convenient to adopt the total gross output of each
industry as the unit of measure of the correspondend commodity;4 the quantity
relations involved in the previous examples thus become:

Example10 : a self-replacing system (unitary gross output)

3/4 unit of wheat  4/13 unit of iron ? 1 unit of wheat
1/4 unit of wheat  9/13 unit of iron ? 1 unit of iron,
= 1 unit of wheat = 1 unit of iron

and

Example 20 : a non-self-replacing system (unitary gross output)

3/4 unit of wheat  2/13 unit of iron ? 1 unit of wheat
2/4 unit of wheat  9/13 unit of iron ? 1 unit of iron,
= 5/4 q unit of wheat = 11/13 unit of iron

and the corresponding price equations become:

    
3=4 4=13 pw p
¼ w ; ð10 Þ
1=4 9=13 pi pi
and
4
This change of the unit of measure allows one to represent technical processes by using production
coefficients without entailing specific assumptions on returns.

123
Econ Polit

    
3=4 2=13 pw pw
¼ : ð20 Þ
2=4 9=13 pi pi

Contrary to systems (1a, 1b) and (2a, 2b), systems (10 ) and (20 ) have different
solutions, because after the normalization adopted for the quantities, the unit of
measure of wheat are 8 qr. of wheat in system (10 ) and 4 qr. of wheat in system (20 )
(the unit of measure of iron remains 13 t. of iron in both systems). Yet, the matrices
involved in systems (10 ) and (20 ) have the same characteristic equation,
52k2 - 75k ? 23 = 0, whose solutions are kM = 1 and km = 23/52: the maximum
rate of profit remains thus equal to zero.
In general, in a system with M industries and commodities, the input matrix and
the output vector are:
2 3 2 3
a11 a12    a1M 1
6 a21 a22    a2M 7 6 17
6 7
A ¼ ½a1 a2 . . . aM  ¼ 6 .. .. .. .. 7 and u ¼ 6 4 .. 7 ; ð3Þ
4 . . . . 5 .5
aM1 aM2    aMM 1

(production processes are represented on the rows). Suppose that matrix A is

indecomposable, that is, all commodities are basic. The productive processes are
represented by the relation:
A ! u:

The ensuing price system is,

Ap ¼ p; ð4Þ
which has non-trivial solutions if det(A - I) = 0, that is equivalent to stating that
the number 1 is eigenvalue of A. The price vector is positive if and only if the
number 1 is the dominant eigenvalue of A,5 that is,
kM ¼ 1: ð5Þ

A system is defined as just viable if it is possible to find a positive vector p which

satisfies (4). Economically, a system is just viable if it is possible to determine a set

5
The Perron-Frobenius theorems state that a semi-positive indecomposable square matrix, A, has an
eigenvalue, kM, with the following characteristics:

i) kM is real and positive and the modulus of all other eigenvalues are lower than kM;
ii) The right-hand and the left-hand eigenvectors associated to kM are positive, while the right-hand
and the left-hand eigenvectors associated to all other eigenvalues have at least one negative
component;
iii) kM is included between the minimum and the maximum of all the sums of the columns (or of the
rows) of matrix A.

Eigenvalue kM is called the dominant eigenvector of A. For details see, for example, Pasinetti (1977a,
Mathematical appendix, § 12.2).

123
 Econ Polit

of prices which puts each industry in the condition to reintegrate the exact value of
the means of production employed. Mathematically, such a vector exists if and only
if the matrix of technical coefficients A satisfies condition (5). On the other hand, a
system is defined as being in a self-replacing state if its technical coefficients matrix
A satisfies the condition:
uT ¼ uT A; ð6Þ
that is, if the sum of each column of input matrix A—which represents the quantity
of each commodity employed as means of production in the various industries—is
equal to 1, i.e. the total output of each commodity. The Perron-Frobenius theorems
[see footnote 5, in particular statement iii)] entail that the dominant eigenvalue of
the technical coefficients matrix of a system in a self-replacing state can only be 1.
Hence, while a system in a self-replacing state is just viable the reverse is not
guaranteed. In fact, given system A ? u, with kM = 1, in general uTA = uT, in the
sense that for some industries uTam \ 1 (surplus industries), but for others uTam [ 1
(deficit industries); condition (6) identifies a subset of matrices among those sat-
isfying condition (5).
There is thus a difference between ‘just viable systems’ and ‘self-replacing
systems’, the former are systems able to determine a set of economically meaningful
relative prices that allows each industry to reintegrate the exact value of the means
of production, the latter are systems where each industry produces a quantity of
commodity equal to the amount required of that commodity as a means of
production by the entire system.6
The case of just viable systems which are not in a self-replacing state is not
excluded by Sraffa, although it is not explicitly considered in the book:7 the only
exception is the footnote appended to the first chapter, which is reproduced in its
entirety here.
This formulation [i.e. the entire Chapter I] presupposes the system’s being in a
self-replacing state; but every system of the type under consideration is
capable of being brought to such a state merely by changing the proportions in

6
In input–output analysis, where the price system is normally paired with the quantity system, the notion
of viability appears essentially as a notion regarding quantities (see, for example, Kurz and Salvadori
1995, chs. 2–4, which develop in detail the case of systems with a surplus). On the contrary, in the present
paper the emphasis is placed on the consequences of the fulfillment of condition (5) for the relations
concerning value. The physical dimension, however, is not lost: (5) is a condition that concerns matrix A,
whose elements are quantities.
7
It is curious to observe that the numerical example provided in Sraffa’s Chapter I cannot be
transformed into a non-self-replacing system through a mere simplification of the price equations by a
whole number, in contrast to our example in Sect. 2: in fact, since both equations in this example,
280pw þ 12pi ¼ 400pw ;
120pw þ 8pi ¼ 20pi ;
can be simplified by the same factor (i.e. by the number 4), we obtain a system which, after
simplification, is still in a self-replacing state:
70pw þ 3pi ¼ 100pw ;
30pw þ 2pi ¼ 5pi ;
where 70 ? 30 = 100 and 3 ? 2 = 5. Is this a coincidence?

123
Econ Polit

which the individual equations enter it. (Systems which do so with a surplus
are discussed in §14ff. Systems which are incapable of doing so under any
proportions and show a deficit in the production of some commodities over
their consumption even if none has a surplus do not represent viable economic
systems and are not considered). (Sraffa 1960, p. 5, fn. 1).
Observe that the definition of the just viable system here proposed is entirely
compatible with Sraffa’s definition of viable systems; that is, systems capable of
being brought to a self-replacing state (or, as will be shown, to a surplus system8) by
changing the proportions in which the equations enter it. In fact, as in viable systems
kM = 1, thanks to the Perron-Frobenius theorems [statement ii)], it is always
possible to find a positive vector, q, such that qTA = qT. Sraffa’s definition of just
viable systems is quite indirect [Chiodi (1992) developed it in detail]. This is
probably due to the fact that Sraffa preferred to avoid to characterize the properties
of his framework in formal terms. Nonetheless, one cannot but admire such an
ingenious definition which enucleates the potential physical aspect of viability.

3 Surplus systems vs viable systems: the case with R > 0

The case considered in the previous Section is limited to the borderline case of a
matrix with a dominant eigenvalue exactly equal to 1 and, consequently, a zero
maximum rate of profit. It was just a simple introductory case. Similar
considerations can, however, be applied to systems characterized by a positive
maximum rate of profit: R [ 0. To obtain an example of such a case, it is sufficient
to reduce the amount of an input in one of the industries provided in Example 2
above. Thus a system able to generate a positive maximum rate of profit is obtained,
but it is not in a self-replacing state. For example, if the amount of iron employed as
a means of production in the iron industry is reduced from 9 to 7, we obtain:

Example 3: a non-self-replacing system

3 qr. wheat  2 t. iron ? 4 qr. wheat
2 qr. wheat  7 t. iron ? 13 t. iron
= 5 qr. wheat = 9 t. iron

The ensuing price system, which must now take into account a uniform positive
rate of profit, R, is:
ð1 þ RÞð3pw þ 2pi Þ ¼ 4pw ; ð7aÞ

8
It is clear from the quoted passage that Sraffa’s definition of viability is not restricted to the case of
systems with zero surplus. We will see in the next section how these arguments extend to the case of
systems with a positive surplus.

123
 Econ Polit

ð1 þ RÞð2pw þ 7pi Þ ¼ 13pi ; ð7bÞ

and its solutions are R = 6.26% and pw/pi = 2.62. Nevertheless, the system is not in
a self-replacing state: there is a deficit of 4 - 5 = - 1 unit of wheat and a surplus
of 13 - 9 = 4 units of iron.
By re-defining the units of measure of each commodity in such a way that the
gross output of each industry is equal to 1, the price system of an economy with M
commodities and M industries becomes,
ð1 þ RÞAp ¼ p: ð8Þ

For simplicity, suppose that matrix A is still indecomposable. System (8) has a
non-trivial and positive solution with respect to prices if and only if R = (1 - kM)/
kM; moreover, the uniform rate of profit is positive if
kM \1: ð9Þ

A system will be called viable if it is possible to find a semi-positive vector, p, and a

positive scalar, R, which satisfies (8). Economically, a system is viable if it is possible
to determine a set of prices which put each industry in the condition to reintegrate the
value of the means of production employed and obtain a uniform positive rate of
profit. Mathematically, this happens if and only if the technical coefficients matrix
satisfies condition (9). Again, viability is not sufficient to guarantee the physical
reproducibility of the entire system. This is ensured by a stricter condition,
uT A  uT ; ð10Þ
which excludes physical deficits in any industry and ensures that at least one
industry produces a positive physical surplus. Condition (10) qualifies a system as a
surplus system. As before, condition (10) entails condition (9)9 but not the reverse.10
It deserves to be noted that the case of systems with a positive maximum rate of
profit (R [ 0) which are not in a self-replacing state probably outlines the most likely
condition of actual economies, with some industries in expansion and other
industries in contraction (see, for example, Pasinetti 1981, ch. V, § 10, or 1993, ch.
IV, §§ 11 and 12). This is the relevant point of the analysis proposed so far: the notion
of production prices is not limited to self-replacing systems (or to growing systems).

9
By result iii) of the Perron-Frobenius theorems, condition (10) entails kM B 1; in order to exclude the
equal sign observe that (10) implies that there is a positive surplus for at least one commodity. Since A is
indecomposable, all prices are positive; thus the value of this surplus is positive, and it is distributed (in
proportion to the value of the means of production of each industry) in the form of profits. Thus the rate of
profit is positive, i.e. R [ 0 and hence kM = 1/(1 ? R) \ 1. In more formal terms, suppose, by
contradiction, that kM = 1; hence, the price vector would satisfy Ap = p which, after being pre-
multiplied by vector uT, entails
uT Ap ¼ uT p: ðÞ
On the other hand, as A is indecomposable, p [ o; post-multiplying both sides of (10) by p obtains
uT Ap\uT p; ð  Þ
but (*) contradicts (**), so the assumption kM = 1 is untenable.
10
As observed in footnote 6, the viability condition (9) is normally derived from the quantity systems;
see, for example, Kurz and Salvadori (1995, Chs. 2–4).

123
Econ Polit

4 Non-basic products and joint production

Till now, the analysis has been carried out for basic commodities only. Nonetheless,
the presence of non-basic products does not alter the conclusions. The numerical
examples provided above may be suitably modified to include a non-basic product
by adding a further process employing one or more basic commodities without
modifying the processes of basic commodities. As is well known, the production
processes of these commodities do not concur to determine the maximum rate of
profit, that depends on the dominant eigenvalue of the technical coefficients sub-
matrix of basic commodities,11 nor do they determine the prices of basic
commodities [for details, see Sraffa (1960, § 6 and Appendix B); see also Varri
(1979)]. Hence, all the previous considerations remain valid: a set of production
processes may be able to determine a uniform non-negative rate of profit and a set of
production prices without being in a self-replacing state. In particular, the
appearance of one or more non-basic products may absorb the means of production
in such a way as to transform a system which is in a self-replacing state as regards
the processes of basic commodities into a non-self-replacing system.
In addition, the distinction between viability and repeatability, as presented in the
previous sections, remains unchanged in the case of joint production. Consider, for
example, the following price system:
ð1 þ RÞð20pw þ 30pi Þ ¼ 50pw þ 5pi

ð1 þ RÞð9pw þ 12pi Þ ¼ 10pw þ 25pi ;

whose solution is R = 50%, pw = 2 and pi = 1. In this case, the system is viable,
yet it is not self-replacing: it has employed 29 units of wheat and 42 units of iron to
produce 60 units of wheat and 30 units of iron; the net products of wheat and iron
are ? 31 units of wheat and - 12 units of iron.

5 Non-self-replacing systems and returns

Non-self-replacing systems have seldom been considered in the classical literature

of production theory and the related literature on input–output models. Several
reasons may explain this lack of interest, although the actual experience of
contemporary economic systems is characterized by the co-presence of declining
and growing industries. One factor which probably has played a relevant role in
disregarding the case of non-self-replacing systems has been the practice of
coupling the analysis of price systems with that of quantity systems. This is typical
of input–output literature, where the attention is mainly put on stationary systems, or
on steady growth systems. There is another reason which discouraged the study of
non-self-replacing systems, which is probably more relevant for economists

11
An exception is represented by the ‘freak’ case considered by Sraffa (1960, p. 25, fn 1 and Appendix
B) of a non-basic product with an own-rate of reproduction so unusually low as to be lower than that of
basic commodities. However, this case is not relevant from the economic point of view.

123
 Econ Polit

interested in Sraffa’s framework: the fact that situations of non-repeatability of

some processes may interact with the issue of ‘returns’. This has always constituted
a tricky point within Sraffian literature. As we have seen, Sraffa defined ‘viable’
those systems which can be put in a self-replacing state by changing the proportions
of the equations entering the system [see the quotation from Sraffa (1960), at the
end of Sect. 2]. A certain amount of turmoil resulted from this definition [see, for
example, Samuelson (2000, §2) and, more recently, Sinha (2015, §2)]. According to
these authors, this would be an operation that requires assuming constant returns to
scale. In his reply to Samuelson, Garegnani—based on the fact that Sraffa speaks of
changes of the proportions among equations and not among industries—suggested
to interpret the changes as virtual, similarly to what Sraffa does in order to build the
Standard system. Considering changes in the proportions among ‘equations’ instead
of among ‘industries’ allows us to imagine hypothetical changes of the scale of
activity of the various industries (instead of actual change of output levels). If these
hypothetical changes of the proportions of the various industries are able to
transform a non-self-replacing system into a self-replacing one then, according to
Sraffa’s definition, it is a viable system because potentially it can be re-
proportioned—under the assumption of unchanged technical coefficients—into a
self-replacing system. In substance, this possibility provides us with a sort of
‘physical support’ of the viability of such a system.
There is, however, another reason to worry about returns in the case of non-self-
replacing systems. In such systems, the quantity produced of one (or more)
commodity is lower than the quantity employed of it as a means of production.
Hence, one or more production processes of the subsequent period must be operated
at a lower scale; perhaps other processes will contemporaneously increase their
scale. These changes of output levels, which take place at each period, may induce a
modification of a set of magnitudes which constitute the data on the basis of which
prices are calculated, that is, technical coefficients. This is an issue which has
already been raised in the literature on production prices. In fact, it goes beyond the
case of non-self-replacing systems as it concerns all cases where changes in output
levels play a preeminent role. Shared conclusions are probably still far from being
reached. Without presuming to definitively settle the question, the identification
carried out here between viable and non-self-replacing-systems forces us to
reconsider the issue of returns in production price systems. The considerations that
will emerge here can be applied to all other cases where output level changes play a
relevant role in the analysis.
We can find at least two general positions on this issue. An attempt will be made
here to evaluate the reasons behind the positions as well as provide a few arguments
in favour of one of them.
According to one view, the price equations are to be conceived as referring to a
time interval containing more than one period of production. The magnitudes
involved in the system, both the unknowns (the prices and the residual distributive
variable) and the data (the quantities produced and employed, that is, the technical
coefficients, and the exogenously fixed distributive variable) are the averages
calculated over such intervals. In this way, growing (declining) industries display an
average output level greater (lower) than the quantity of that commodity employed

123
Econ Polit

as means of production. According to this view, the technical coefficients involved

in the price equations are the coefficients that are observed on average over that
interval. There is, thus, no need to assume their constancy during the entire interval.
It is sufficient to require that the changes are not too large and frequent, for example,
una tantum changes, or that the data change at a slower pace with respect to the
speed by which the system converges to its normal position, so that the latter is not
significantly modified when the system gravitates around it.
This view has the merit of avoiding any assumption on the constancy or the
variability of technical coefficients,12 as well as that of analysing the phenomena of
the gravitation around normal positions and the modification of the centres of
gravitation within the same framework. But, if one wanted to zoom in on the
workings of these phenomena and analyse how the sequence of production periods
unrolls as time goes by,13 it is no longer possible to adopt the device which
considers ‘[n]o changes in output and … no changes in the proportions in which
different means of production are used by an industry’ (Sraffa 1960, p. v), because
in these cases, attention is placed exactly on the change of output levels occurring at
each production period. In this perspective, production prices lose their persistence
as long as changes in output levels affect technical coefficients. This may be the
case in order to undertake a step by step analysis of the sequence of production
periods of non-self-replacing systems, but the same situation could happen by
looking at the periodical sequence of a (uniform or non-uniform) growing system, as
well as in the gravitation process around the normal position. In any of these cases,
if production prices are to be regarded as centres of gravitation, one must allow that
transactions are repeated in each production period on the basis of (nearly)
unchanged data (see Garegnani 2007, pp. 228–229). And the only case where the
same price system survives for more than one period of production is when
technical coefficients remain constant,14 at least for comparatively small changes in
output levels. The same assumption is to be made in growing systems, or in the
analysis of the gravitation process in order to avoid that such a process entails some
form of path dependency: as capital mobility implies actual changes of sectoral
output levels it must be assumed that technical coefficients remain constant,
otherwise the normal position around which the system gravitates would depend on
the actual path followed by the output levels of the system.15
The assumption of technical coefficients constant with respect to changes in
output levels represents a discontinuity with respect to the line followed by Sraffa.
12
For appreciating the reasons why Sraffa could avoid to introduce assumptions on returns, see Eatwell
(1977).
13
It is to be noted that the gravitation process is supposed to take place in real time (not in virtual time,
as with Walrasian tâtonnement) through the movements of capital among industries. Consequently, if
capital has to move from one industry to another, it must remain employed in each industry for at least
one production period. The achievement of a long period position, characterized by uniform rates of
profit, thus takes more than one production period.
14
A similar point is discussed in Chiodi (1998, in particular § 4).
15
On this see, for example, Garegnani (1990), where the same technical coefficients are employed in the
system defining the ‘natural prices’—his Eq. (1) —and the equations defining the ‘market rates of
profits’—his Eq. (3)—which clearly refer to actual sectoral output levels, that in general do not coincide
with the ‘normal’ configuration.

123
 Econ Polit

Therefore, it becomes necessary to fully understand the reasons why Sraffa was so
careful in keeping his analysis free from assumptions on returns, in order to asses if
this specific assumption on technical coefficients can be adopted without
prejudicing the goals pursued by his analysis.
It is possible to say that the main purpose pursued by Sraffa in his book is to build
a coherent theoretical framework of the relations between the prices of commodities
and the distributive variables for a system where commodities are produced by
means of commodities and labour. To this purpose it is necessary to avoid any
connection with notions like demand and supply functions, or preferences and factor
endowments. Demand functions and/or preferences would introduce those subjec-
tive elements into the theory which Sraffa preferred to avoid;16 moreover, the
discovering of phenomena like the reswitching of techniques and reverse capital
deepening would have led to factor demand functions with unconventional slopes.
Supply functions and endowments, on the other hand, would have clashed with the
impossibility of measuring the endowment of one factor, that is, capital. Sraffa’s
price system is built in such a way as to avoid all these ‘dangerous’ links. Its main
features can be outlined as follows:
Proposition 1 (Separate determination of prices and distribution from output
levels) In a single product circular production system, the composition and the level
of the social product and one distributive variable are determined separately from
the prices of commodities, and are therefore taken as given when writing the price
equations of these commodities.
This peculiar result of Sraffa’s framework reflects the methodology typically
followed by classical economists [a detailed description of the logical framework of
classical economists is provided by Garegnani (1984)]. The result is compatible
with a vision of the process of income distribution as essentially based on
institutional circumstances. On the contrary, in the neoclassical approach, the link is
univocal: commodity prices as well as distributive variables reflect factors
endowments and consumers’ preferences: each factor of production is ‘priced’ by
the market according to its availability and by the demand for the final
commodity(ies) it contributes to produce. The higher the proportion in which a
final good is demanded, the higher the remuneration received by those factors which
are more intensively used in its production. Within such a framework, at least in its
simpler and more abstract version, there is no space for actions to modify income
distribution (i.e. redistributive policies).
The fact that classical economists have never explicitly outlined an analytical
link between commodity prices and the residual distributive variable with individual
preferences and endowments has always been seen as a deficiency in the eyes of
neoclassical economists. It constitutes an exception to the usual way to solving the
problem of the choice of technique in general equilibrium analysis, where
preferences and endowments, that is, relative scarcity of factors, do affect the
technique adopted and, consequently, commodity prices and income distribution.
The main features of such a sub-case have been in fact codified into a ‘non-

16
On this see Kurz and Salvadori (2005).

123
Econ Polit

(a) (b)
Fig. 1 Marshallian view of price determination

substitution’ theorem, that holds when cost curves are horizontal, that is, under
(Marshallian) constant returns. The statement of this theorem is:
Proposition 2 (Non-substitiution) In a single product circular production system,
if there are one non-produced factor of production and constant returns to scale,
once one distributive variable is fixed exogenously with respect to the price
equations, the technique adopted, the prices of commodities and the other
distributive variable are independent of the level and the composition of final
demand.
The first explicit statement of this Proposition as a theorem was provided by
Georgescu-Rogen (1951) and Samuelson (1951). However, its essence can be easily
grasped by a comparison of a couple of Marshallian supply and demand graphs (see
Fig. 1).
The case of constant unitary costs (i.e. constant marginal and average cost
curves) is represented on the right-side of the diagram. Marshall and several
neoclassical authors saw it as the junction of the marginalist and the classical
theories of value: the classical theory would appear, thus, as a particular case of the
marginalist theory.17

17
The discriminatory element of Fig. 1b with respect to Fig. 1a is the constant supply curve. This
requires:
M1) Constant returns to scale;
M2) Constant rewards of non-produced factors as long as the output level changes.
The same situation is guaranteed in the non-substitution theorem by assuming:
S1) Constant returns to scale;
S2) The existence of just one primary (scarce) factor, so that factors’ rewards cannot express their
relative scarcity;
S3) A given rate of interest.
Assumptions M1 and S1 are equivalent; assumptions S2 and S3 entail assumption M2.

123
 Econ Polit

The difference in the formulation of Propositions 1 and 2 is crucial in

understanding the different meaning they aim to convey and the reasons which
support them. The so-called ‘non-substitution’ result, either in its Marshallian
representation or in the formulation proposed by Samuelson, is the analytical
consequence of the assumption of a constant unitary cost curve or of the equivalent
set of assumptions: constant returns to scale, a single primary factor and a given rate
of interest (see footnote 17). A relaxation of one of these assumption immediately
re-establishes the full dependence of prices and distribution on technology,
individual preferences and endowments. The separate determination of output levels
and of one distributive variable from commodity prices and the residual distributive
variable that are found in classical economists reflects the vision that income
distribution is essentially a social matter, rather than the result of the ‘composition’
of the interests of ‘rational’ individuals carried out by the ‘market’. The classical
outline of the distributive phenomenon needs no particular assumption on returns, as
in the determination of commodity prices and of the residual distributive variable,
output and input levels are taken as given. This explains why Sraffa places such
emphasis on the fact that his analysis is free of any assumption on returns:18 he
wants to stress the generality of his results and prevent his theory as being regarded
as a particular case of general equilibrium analysis.

18
Similarly, Sraffa vehemently rejects the idea that his prices are independent of demand. In a famous
letter addressed to Arun Bose (classified in Sraffa’s archive as SP, C32/3), Sraffa writes:
Trinity College,
Cambridge.
9th December, 1964.

Dear Arun,

I am sorry to have kept your MS so long - and with so little result.

The fact is that your opening sentence is for me an obstacle which I am unable to get over. You
write: ‘‘It is a basic proposition of the Sraffa theory that prices are determined exclusively by the
physical requirements of production and the social wage-profit division, with consumers demand
playing a purely passive role.’’

Never have I said this: certainly not in the two places to which you refer in your note 2. Nothing, in
my view, could be more suicidal than to make such a statement. You are asking me to put my head
on the block so that the first fool who comes along can cut it off neatly.

Whatever you do, please do not represent me as saying such a thing.

This initial and to me quite maddening obstacle has prevented me, in spite of many attempts, from
reading understandingly your article. You must find a more detached reader to advise you about it.
I am very sorry to seem so unhelpful, but I have spent quite a lot of time upon your work, to no
purpose. I do not think that it would be any good keeping it longer, so I now return it to you.

Yours sincerely,
By the way, without constant returns, a change in the level or in the composition of final demand does
affect production coefficients, so that prices as well as the rate of profit change; but, these effects have
nothing to do with the monotonic relations between the proportions in which final goods are demanded
and the prices of the factors more intensively employed in producing those final goods [on this, see
Pasinetti (1977b)].

123
Econ Polit

Now, let us return to the original matter, that is, the legitimacy of assuming the
constancy of technical coefficients with respect to changes in output levels. The
coherence of Sraffa’s analytical framework and, in particular, the choice of
determining one distributive variable outside the price equations—that is, the
pure institutional character of the phenomenon of income distribution—is not
questioned at all by the assumption that technical coefficients are constant with
respect to changes in output levels. Clearly, this assumption reduces the generality
of the analysis, but it is needed to keep the normal price configuration sufficiently
persistent in order to be regarded as a centre of gravitation if our investigation
explicitly aims to describe the sequence of production periods. It can be added that
for our purpose it is sufficient that the constancy of technical coefficients holds for
contained changes in output levels: it is plausible that comparative small changes in
output levels give rise to negligible changes of technical coefficients. This was the
case, for example, for the VHS recorder industry: the reduction in output took place
through a progressive dismantling of plants devoted to this production line; there is
no reason to imagine that such a change have entailed a change of technical
coefficients. On the contrary, significantly large changes in the output level of one or
more commodities probably impose a change of technical conditions and thus of the
centre of gravitation; but, this is a phenomenon that must be studied outside the
method of normal positions.

6 Concluding remarks

The distinction between viable systems and self-replacing systems here proposed
has revealed a fundamental characteristic of the notion of prices determined in
Sraffa’s framework: they are those prices that, if adopted, allow each industry to
recover the expenses of production and obtain a uniform (non-negative) rate of
profit [on this see also Piccioni (2000, p. 192)]. It has been shown that production
prices do not presuppose that the entire system is in a self-replacing state. In fact,
production prices can be computed also for non-self-replacing systems. In other
words, within Sraffa’s framework, viability is an issue that is better appraised if it is
connected to the value side, while the conditions of self-reproduction pertain to the
quantity side. In the approach adopted by Sraffa, where quantities are taken as
given, this distinction must be taken into consideration.19 This distinction and its
analytical characterization constitute one goal of the present paper.
19
This is a distinction that releases Sraffa’s prices from any direct link to an equilibrium between
quantities: Sraffa’s prices do not originate from equality between demand and supply. Thus, it seems of
no immediate utility to equip Sraffa’s price equations with a set of quantity equations like in the von
Neumann model, where one set of variables is ‘dual’ to the other. This does not prevent that there are
situations where Sraffa’s price equations can usefully be paired with a set of conditions involving outputs
and quantities demanded. For example, this is the case of formal models of gravitation, where the long-
period position is characterized by:

a) equality between the output level of each commodity and the corresponding demand,
b) the possibility of indefinitely replicating the same output levels (steady state) or increasing these
levels, leaving the proportions among industries unchanged (balanced growth).

123
 Econ Polit

The analysis carried out of non-self-reproducing systems has led to the question
of how to handle cases of systematic changes in output levels in order to maintain
the property of persistency for production prices, which is the basis for interpreting
them as centres of gravitation. Two stances can be adopted. If production prices are
referred to a time interval during which moderate changes in output levels occur,
one can regard the data entering the price equations (i.e. the quantities of
commodities produced and employed as means of production) as the averages of
such magnitudes over these intervals. If, alternatively, the aim is to detail the
sequence of the various production periods, it is reasonable to assume that
production coefficients remain constant, at least for comparatively small changes in
output levels.20 It has been argued that this does not conflict with the targets pursued
in Sraffa’s book. The ultimate reason why Sraffa wanted to keep his analysis free
from assumptions concerning returns was to avoid his framework from being
perceived as a particular case of a general equilibrium system, where constant
returns, together with an artificial exogenous fixation of one distributive variable,
makes prices and the remaining distributive variable independent of final demand.
However, once it is acknowledged as a typical feature of the classical approach the
fact that output levels and one distributive variable are determined separately from
the relations bearing commodity prices and the residual distributive variable, it
becomes evident that the assumption that technical coefficients are constant with
respect to changes in output levels does not undermine the logic and the objectives
of Sraffa’s reconstruction of the classical surplus approach to value and distribution.
This is the second goal of the present paper.

Acknowledgements I wish to thank Christian Bidard, Antonia Campus, Guglielmo Chiodi, Roberto
Ciccone, Saverio Fratini, Heinz Kurz, Sergio Levrero, Sergio Nisticò, Sergio Parrinello, Luigi Pasinetti,
Fabio Ravagnani, Andrea Salanti, Neri Salvadori, Paolo Trabucchi and Paolo Varri for the discussions on
the topics here presented. I also thank three anonymous referees of this Journal for their comments and
criticisms to a previous version of this paper. Finally, I am grateful to Micaela Tavasani for revising the
English.

References
Cesaratto, S. (1995). Long-period method and analysis of technological change: Is there any
inconsistency? Review of Political Economy, 7(3), 249–278.
Chiodi, G. (1992). On Sraffa’s notion of viability. Studi Economici, XLVII(46), 5–23.

Footnote 19 continued

But, as regards (a), the main force pushing actual (‘market’) prices toward production prices and, thus,
the rates of profit toward a uniform level, is competition among capitalists, i.e. capital movements in
search of the highest rate of profit, not a tâtonnement process driven directly by the excesses of quantities
demanded with respect to the quantities supplied. As regards (b), the representation of the long period
position as a steady state or a balanced growth path is merely a simplification acceptable as long as the
process of structural change takes place at a slower pace with respect to capital mobility induced by the
differences in the rates of profit of the various industries [on this see Cesaratto (1995), and Duménil-Lévy
(1995)].
20
A similar position seems to be expressed by Ravagnani (2001, p. 361, fn. 2).

123
Econ Polit

Chiodi, G. (1998). On non-self-replacing states. Metroeconomica, 49(1), 97–107.

Duménil, G., & Lévy, D. (1995). Structural change and prices of production. Structural Change and
Economic Dynamics, 6(4), 397–434.
Eatwell, J. (1977). The irrelevance of returns to scale in Sraffa’s analysis. Journal of Economic
Literature, 15(1), 61–68.
Garegnani, P. (1984). Value and distribution in the classical economists and in Marx. Oxford Economic
Papers, 36(2), 291–325.
Garegnani, P. (1990). On some supposed obstacles to the tendency of market prices towards normal
prices. Political Economy—Studies in the Surplus Approach, 6(1–2), 329–359.
Garegnani, P. (2007). Professor Samuelson on Sraffa and the classical economics. The European Journal
of the History of Economic Thought, 14(2), 181–242.
Geortescu-Rogen, N. (1951). Some properties of a generalized Leontief model. In T. C. Koopmans (Ed.),
Activity analysis of production and allocation (pp. 165–173). New York: Wiley.
Harrod, R. F. (1961). Production of commodities by means of commodities. Prelude to a critique of
economic theory. The Economic Journal, 6(284), 783–787.
Kurz, H., & Salvadori, N. (1995). Theory of production—A long-period analysis. Cambridge: Cambridge
University Press.
Kurz, H., & Salvadori, N. (2005). Representing the production and circulation of commodities in material
terms: On Sraffa’s objectivism. Review of Political Economy, 17(3), 69–97.
Pasinetti, L. L. (1977a). Lectures in the theory of production. London: Macmillan.
Pasinetti, L. (1977b). On non-substitution in production models. Cambridge Journal of Economics, 1(4),
389–394.
Pasinetti, L. L. (1981). Structural change and economic growth—A theoretical essay on the dynamics of
wealth of nations. Cambridge: Cambridge University Press.
Pasinetti, L. L. (1993). Structural economic dynamics—A theory of the economic consequences of human
learning. Cambridge: Cambridge University Press.
Piccioni, M. (2000). Prodotti netti negativi, accumulazione di capitale e cambiamenti nelle condizioni
tecniche di produzione. In M. Pivetti (Ed.), Piero Sraffa—Contributi per una biografia intellettuale
(pp. 190–202). Torino: Carocci.
Ravagnani, F. (2001). Notes on a mischaracterization of the classical theory of value. Review of Political
Economy, 13(3), 355–363.
Samuelson, P. A. (1951). Abstract of a theorem concerning substitutability in open Leontief models. In T.
C. Koopmans (Ed.), Activity analysis of production and allocation (pp. 142–146). New York: Wiley.
Samuelson, P. A. (2000). Sraffa’s hits and misses. In H. Kurz (Ed.), Critical essays on Piero Sraffa’s
legacy in economics (pp. 111–180). Cambridge: Cambridge University Press.
Schefold, B. (1989). Mr Sraffa on joint production and other essays. London: Unwin Hyman.
Sinha, A. (2015). A reflection on the Samuelson-Garegnani debate. Economic Thought, 4(2), 48–67.
Sraffa, P. (1960). Production of commodities by means of commodities—Prelude to a critique of
economic theory. Cambridge: Cambridge University Press.
Sraffa, P. (1962). Production of commodities: a comment. The Economic Journal, 72(286), 477–479.
Varri, P. (1979). Basic and non-basic commodities in Mr. Sraffa’s Price System. Metroeconomica,
XXXI(1), 57–72.

123