INTRODUCTION

^{Steedman (1975}) gave an example of a pure joint production system (i.e.,, without fixed capital or land) in which the rate of profits and their associated prices of production were positive and yet aggregate “surplus value” was negative. This example sparked a controversy related to the questions of how to define labour values in the context of joint production and particularly whether or not it implied a refutation of the so-called “fundamental Marxian theorem”. This “theorem” states that a positive “rate of surplus value” - the ratio between the quantity of embodied labour on the physical surplus (“surplus value”) and the quantity of embodied labour in the necessary consumption of workers (“variable capital”) - is a necessary and sufficient condition for a positive general rate of profits.

Although the literature on the possibility of *individual* negative labour-values quite exhaustively scrutinized the necessary and sufficient conditions for it to happen - see ^{Schefold (1989}), ^{Filippini & Filippini (1982}) and ^{Hosoda (1993}) - the same cannot be said on the literature related to the possibility of negative surplusvalue - i.e.,, the labour-value of the aggregate surplus product of the economy. Many authors tried to get around Steedman’s critique through redefinitions of the labour values of single commodities, arguments about heterogeneous labour and considerations about the problem of choice of technique. Few contributors, however, discussed the actual economic meaning and relevance of the special (implicit) assumptions in Steedman’s example.

In this paper we do that by means of a critical survey and a theoretical evaluation of this literature^{1}. We make use of Steedman’s original numerical example as basis to compare and contrast the different contributions using the activity analysis diagram^{2} which provides a very clear picture of some properties of a square twocommodity system^{3}.

Our main conclusions are that although labour values for some single commodities can indeed be negative in the most general case (it *may* happen if the system is not “all-productive”, as defined by ^{Schefold (1978})), the basic and sensible classical proposition that in the aggregate less labour would be necessary to produce only the necessary wage basket than to meet also the total final demand coming from the expenditures of the capitalists always holds, even if in some cases (as in the particular economic system depicted by Steedman) producing the necessary wage basket could entail also jointly producing some extra outputs unneeded by the workers.

Besides scrutinizing the original example and the controversy, we also propose a reformulation of a method originally presented by ^{Akyüz (1983}) and with that it will be shown that it is always possible to find an economically meaningful amount - i.e.,, in the precise sense that contains only feasible (non-negative) levels of activity - of aggregate surplus-value for the economy without changing the processes in use (i.e.,, using the dominant techniques in use in the square system). As it will be seen, this method seems to be the only available one which is coherent with the classical-marxian idea of taking as given the processes of production in use to measure the aggregate amount of extracted surplus labour.

The paper is organized as follows. Second section briefly discusses issues related to labour values in the context of joint production since the original contribution of ^{Sraffa [1960}]. Third section presents and discusses Steedman’s assumptions and results. Fourth section reviews the debate on positive profits and negative surplus value. Fifth section offers concluding remarks.

LABOUR VALUES AND JOINT PRODUCTION

Labour values, in the case of homogeneous labor, are the physical quantities of labour directly and indirectly necessary to produce a unit of a particular commodity using the dominant (or socially necessary) methods of production actually in use. ^{Sraffa (1960}) has shown, for single production systems, that the set of prices of production measured in labour commanded (i.e.,, divided by the money wage) coincide with the set of labour values when the rate of profits is zero.

In the general case of joint production the author points out that there is naturally an obvious difficulty in thinking of what is the quantity of labour directly and indirectly necessary to produce a particular single commodity, since more than one commodity can be produced by the same process and at the same time that same commodity may be produced by other processes of production.

For the cases where it is possible to produce single commodities separately, i.e.,, to increase the net output of a particular commodity without necessarily increasing the net output of any other^{4}, it is also possible to calculate precisely what is the quantity of labour directly necessary to produce only one unit of that particular commodity and, thus, there is no difficulty in calculating its positive and additive labour value. That seems to be why simple systems with non-shiftable fixed capital or the analysis of land and differential rents tend to behave like single product systems. In general the necessary (but not sufficient) condition for non separability is that there are processes in use that produce net outputs of more than one commodity.

However, things get more complicated within pure joint product systems that do not possess the property of separability. The difficulties with the concept of labour value of individual commodities arise exactly because of this reason in these systems.

For these type of systems ^{Sraffa (1960}, p.59) mentions the possibility that the labour values of individual commodities could turn out to be negative. The author explains the meaning of such negative labour values by referring to the fact that labour values are one and the same thing as the employment multipliers of a single commodity associated to that system of production. A negative labour value thus means that if we think of increasing the net product of only that particular commodity by one unity, and production is not separable and thus we will also increase the net output of at least one other commodity, we will necessarily have to increase the amount of labour employed by one process but also will have to decrease the level of employment in at least some other process to prevent the overproduction of the second commodity.

Now depending on the direct amount of labour employed in the process that is expanded compared with the process that is contracted, it may happen that in the end the total amount of labour employed in all processes will be lower than it was before the increased of the net output of the first commodity. In this case the commodity in question will have a negative labour value because an increase in its production has led to a decrease in total (direct and indirect) employment.

^{Sraffa (1960}) also warned in a footnote^{5} of the possibility of the awkward occurrence of what he called “negative industries” by which he meant that the fact that since under joint production the activity levels of some processes must be decreased when that of the others increase to match the level of composition of the “requirements of use” was logically possible, if the change in the level and structure of demand was sufficiently drastic, that the contraction of some particular joint processes of production could be so extreme as to require it to be “activated” at a negative level. Since negative activity levels do not exist, this only means that in fact that even contracting these processes to zero the net output of the overabundant products would not be reduced sufficiently for the system to adjust the vector of production exactly to the vector of effectual demands - in this case, the processes in use would have to change in order to match exactly the vector of effectual demands.

STEEDMAN’S “POSITIVE PROFITS WITH NEGATIVE SURPLUS-VALUE” PARADOX

^{Steedman (1975}) seems to have taken this route and, by inverting the assumptions that Sraffa used to rule out processes with negative activity levels, produced his own particular example. The inconclusive results of the controversy sparked by his “positive profits with negative surplus value” example seem to have been due to the combination of two main factors. The first is that neither much explanation of the economic meaning of the original results nor of the relevance of its assumptions has ever been made^{6}. The second reason is that, with very few exceptions, his critics seemed so concerned to “defend” the concept of surplus value and/or the labour theory of value in some form that they tended mostly to add further ad hoc assumptions to those of Steedman’s example in order to change its conclusions, instead of scrutinizing the nature of the example and its implications.

^{Steedman (1975}, p. 115) assumes an economy capable of producing a surplus of two commodities using two joint production processes that use only circulating capital:

INPUTS | OUTPUTS | |||||||
---|---|---|---|---|---|---|---|---|

commodity 1 | commodity 2 | labour | commodity 1 | commodity 2 | ||||

5 | ⊕ | 0 | ⊕ | 1 | → | 6 | ⊕ | 1 |

0 | ⊕ | 10 | ⊕ | 1 | → | 3 | ⊕ | 12 |

He also assumes that the wage is paid at the end of the production period and that the real wage basket is exogenously given: b = (0.5, 0.83). This assumption is not necessary for the present purposes - it does not matter for the results if wages are paid ex-ante or ex-post.

The really crucial assumptions implicit in his analysis amount to three:

(i) A two commodities square system in which both processes generate the same joint products.

(ii) There is a process which is strictly superior to the other one, the second one, having higher net products for both goods

^{7}. In Steedman’s original example we have the following net products of each process (operated at unity levels): m_{1}=(1, and m_{2}=(3, 2), the columns of the matrix (B-A).

These two assumptions imply that the *individual* labour value for the good one is negative. The third assumption is:

(iii) The proportions in which commodities one and two are demanded by the capitalists are very different from the proportions in which they appear in the wage basket.

This will mean that no combination of the two processes available is capable of producing without overproduction either only the wage basket or only the profit earner´s basket. As we shall see it is this latter assumption of the unfeasibility of producing the baskets of the two classes separately plus the great divergence among the bundles, not the mere existence of a negative labour value for one of the commodities, which is the key for the occurrence of negative *aggregates* of labour.

^{Steedman (1975}, p. 115) first shows that in this particular joint production system the rate of profits and relative prices of production are positive, since the profit rate depends only on the existence of a positive net product (surplus product) after the deduction of the necessary consumption for being positive - independently of the labour value measurement.

He then shows that one of the two commodities has a negative labour value:

Where λ_{1} and λ_{2} are the labour values of each commodity. The solution of this system gives us the following vector of labour values:

Where a_{L} is the vector of direct labour requirements (composed only by units, given the normalization presented in the table above) and Λ is the vector of labour values (which also corresponds to the relative prices associated to a profit rate equal to zero, using the wage rate to normalize them).

He then moves on to show also that the aggregate amount of surplus value in this system is negative, both measured as the aggregate labour value of the sum of commodities demanded by the capitalists and as the difference of total labour employed versus the aggregate labour employed (embodied) in the wage basket.

The total net product is y is the sum of each class bundle y_{k} = (5, 2) which is the final demand which is appropriated by the capitalists, and, y_{w} = (3, 5), the workers’ consumption.

To produce this final demand

a feasible combination of both processes is required:

Where x_{1} and x_{2} are the levels of activity of each process required to produce the net product (the components of the vector of activity levels). The solution of this system gives the following vector:

The aggregate amount of labour is given by the sum of each element of the activity level vector using this normalization:

If we multiply the final demand baskets of each class by the labour values of each commodity, this particular technology and pattern of final demand will give us the following labour value aggregates:

Where S is the surplus value and V is the variable capital. The other way for calculating the surplus value is looking at the difference between the living labour (6 units) and the variable capital (7 units):

Before moving on, we think that we can understand this result better if we make the calculation in terms of activity levels associated to the production of each bundle, instead of just finding the labour value of each bundle, as we have done for the whole net product above. For the workers’ bundle

we need to solve:

Which gives us the following “activity levels” required to produce workers’ consumption:

The sum of each component of x_{w} gives us the variable capital. Applying the same procedure for the capitalists’ bundle we get:

Thus, the surplus value S can be calculated from the sum of the components x_{k}.

This way of calculating shows that both bundles require negative activity levels of one of the processes because both bundles cannot be produced *separately* by the processes in use.

Grafically, we have that:

The dotted lines provide the set (“cone”) of feasible net products using the two processes given by the vectors m_{1} e m_{2}. The horizontal axis represents the net product of first commodity and the vertical axis the the net product of the second commodity.

As we can see, the total final demand y falls inside the cone, which means that it can be reached with non-negative levels of activity. The total employment required for that is 6 units. However, the bundles that each class receives fall outside the cone. This means that they cannot be produced *separately* - though it is feasible to produce both jointly. The consequence of this is that negative levels of activity would be required to produce only workers’ consumption or only the final demand of capitalists. If both bundles were inside the cone no negative surplus value could happen, although the labour value of one of the two commodities would still be negative. Indeed, with the same aggregate final demand y=(8, 7) but with the consumption of the workers being y_{w}=(3, 2.5) and the expenditures of the capitalists being y_{k} =(5, 4.5), for example, we would still get, overall activity levels x=(5, 1) and total employment L=6. But now x_{k}=(3.5, 0.5) and thus S= 3.5+0.5=4 and surplus value would be positive. For the workers we would have x_{w} = (1.5, 0.5) and V=2 and thus S= 62=4.

In the example, for the capitalists’ bundle we have that the total employment required to produce it is negative, i.e.,, negative surplus value. So, we may say that within the group of final demands that are non-feasible, there is a group which “requires” negative amount of employment to be produced, which in this numerical example^{8} are those ones that lie below the line y_{2}=0,5y_{1}. But the central point we want to emphasize is that any final demand outside the cone is economically meaningless^{9}.

We can thus see that the existence of negative labour values for single commodities does not by itself imply the paradox. The paradox of positive profits and negative surplus value provided by ^{Steedman (1975}) rests on the fact that negative levels of activity would be required if the bundles were to be produced separately - or, that these bundles can only be produced jointly within this square system.

Of course negative activity levels do not exist. They are just the mathematical symptom of the fact that there would be overproduction of one of the two commodities if we were to produce only the wage bundle of the workers using these two processes. So the right conclusion from Steedman’s example is that in general joint production systems the calculation of aggregate surplus labour should be adapted to deal with the possibility of notional overproduction, instead of using the economically meaningless notion of negative activity levels.

SURPLUS VALUE AFTER STEEDMAN

An initial reaction to the Steedman’s example came from ^{Morishima (1976}). Based on his earlier works such as ^{Morishima (1973}) and ^{Morishima (1974}), the author proposed to redefine the labour values in joint production using what he calls the “true values”. The “true value” of a commodity (or a bundle) is given by the minimization of labour-time required to produce one net unity of it. Be e_{i} the column vector in which the i-th coordinate is unity and the other ones are null, the true-value of the commodity i will be given by the following minimization problem:

Where activity levels and the processes in use are the endogenous variables of the problem^{10} and the symbol ≥ means that the vector is equal (in every coordinate) or higher (in at least one coordinate) than the other one. To find the “true variable capital” e_{i} must be substituted by y_{w}, which is going to be the total employment required to produce workers’ consumption which minimizes the amount of labour expended.

Be x_{w}
^{*} the activity levels that solves the problem, the “true variable capital” would be:

And the “true surplus value”:

Using Steedman’s example, graphically it would be:

Where y_{w}
^{*} is the net product associated with the level of activity x_{w}
^{*}.

In this case, only the process with higher productivity would be operated (this is why y_{w}
^{*} lies above the same line as m ). Because operating this process alone cannot produce exactly the workers’ consumption, there would be excess production of the first commodity of 4,5 units. But the point is that less labour than the total employment would be required to produce workers’ consumption. To produce y_{w}=(3, 5) it would be necessary to operate only the second processes with 2,5 units of labour. This gives a positive “true surplus value” of 6-2,5=3,5.

The criticisms of this procedure are not new: “true-values” are not additive like in Marx, i.e.,, the “true-value” of a bundle of commodities is not equal to the sum of “true-values” of the same commodities separately produced. The authors argue that this different definition of labour value would have a textual basis in Marx but the argument is not very convicing (^{Steedman, 1976a}). A second important criticism is that in this case the (redefined) surplus value would be related to a non-capitalist (i.e.,, nonprofit-maximizing) criterion for the choice of technique - while in Marx and in the literature related to the “fundamental marxian theorem” this was based on the processes in use in capitalism (“the socially necessary techniques”) (^{Akyüz, 1983}). In fact the “true value” calculations measure the hypothetical minimum amount of labour that would be necessary to produce the wage basket for society as such and not the amount that is “socially necessary” given the techniques and processes actually already chosen by a capitalist criterion of choice of technique.

^{Wolfstetter (1976}) questioned Steedman’s use of Sraffa’s approach of assuming square joint production systems with equalities, instead of starting from Von Neumann’s method of rectangular systems and using inequalities. ^{Steedman (1976b}) promptly replied that in the particular case of his example the difference between the two methods would hardly matter for his results, the solution being exactly the same for both approaches. Moreover, Wolfstetter’s third theorem (^{Wolfstetter, 1976}, p. 867), which states that the existence of an inferior process is a necessary and sufficient condition for negative individual labour values, is correct only in a two commodity system as pointed out in the footnote 3 above.

^{Kurz (1979}) and ^{Krause (1980}) tried to refute the paradox redefining the vector of individual labour values to make it semi-positive. ^{Krause (1980}) argued that the different labour productivities are a case of heterogeneous labour and with this argument the author changes the weights of the direct labour vector in such way to guarantee strictly positive individual labour values. The obvious problem with this complicated solution is that the assumption of heterogeneous labour is simply not present in ^{Steedman (1975}). ^{Flaschel (1983}, p. 437) adds relative prices to the definition of individual labour value in order to get positive magnitudes for the latter. This procedure goes in a completely different route with respect to the original function of the labour values of being a measure of physical costs, independent from (and a major determinant of) income distribution and relative prices.

^{Akyüz (1983}) argues that we can always find yw that is a fraction of the actual net product of the economy in such a way that it is enough to achieve the actual workers’ consumption. That is, some measure of the ‘variable capital’ using the processes actually in use and keeping the same proportion in which they are actually operated can be found and used as the basis to calculate the surplus-value of more general joint production systems. He adds that using this method there’s no need for solving a minimization problem and that this would be a great virtue of it (Akyüz, 1982, p. 171).

The problem is that in this case there are many scalars that satisfy the inequality and we would have an indeterminacy problem. It seems to us that the author’s insight can be improved if presented as a minimization problem of finding the minimum scalar that satisfies the above conditions in the following way. This notional scale of production should be given by a scalar Φ given by the following minimization problem:

Where y is the actual net product of the system and yw is the actual workers’ consumption basket. Graphically, we would have:

As we can see that the vector falls inside the cone and, consequently, a positive amount of employment is required to produce it. In this case, we would have a redefined vector of workers’ consumption which has the advantage of being based in the processes actually in use and keeping the same proportion in which they are actually operated. This redefined surplus value S’ will be given by:

If we make a comparison with the “true-values” we can see that this method requires much less new information and is much closer to the actual system data than the former. In the procedure proposed by ^{Morishima (1976}) the processes and their relative levels of activity are altered (besides the presence of non-profit maximizing criterion for the choice of technique). In the method proposed by ^{Akyüz (1983}) there’s only the scalar as a novelty and no change in the choice of techniques is required. Inevitably (as in the “true-value” approach) there’s some excess of production - of the first commodity (2,174 units in the example).

Our improvement to Akyüz’s method allows us to see that the correctness of the sensible classical proposition that (without changing the actual processes that are in use) in a capitalist economy workers obviously work more than they would need to produce only (at least) what is contained in their wage basket.

This proposition relating positive profits and positive surplus labour is of general validity with or without joint production, and whatever may be happening to the labour value of individual commodities. It is important to point out that this “notional overproduction” is not something that will or even may actually happen in this economy. The term means only that this excess would happen if the system were to produce these different quantities using the same techniques. But this is not the case and these calculations are just like Sraffa’s standard system: abstractions that reveal some important properties of the actual system - in our case, the possibility of finding an economically meaningful measure of aggregate surplus value. The calculations only provide a useful piece of information: how much labour would be needed to produce with minimum waste the workers’ consumption basket.

More recently ^{Trigg and Philp (2008}) have claimed that there will always be positive surplus value in the aggregate - even in the presence of individual negative labour values for single commodities - using the argument that Kahn’s multiplier is positive and it depends only on the fact that the value of the labor power is below one and this will always be true for systems that produce positive surplus product. In order to understand their point, we have to start with the gross product decomposition

This can be rewritten in the present case as

Which can be rewritten as

Pre-multiplying both sides by the labor-value vector we get

Thus,

Thus, from a Kahn-type multiplier perspective, we can see the equation in the following way: total employment is a multiple of the employment required to produce capitalists’ final demand. This multiple - the Kahn “multiplier” - depends on the value of the labour power (Λb in the denominator). Thus, the authors argue, it is enough the value of labor power is smaller than one for the multiplier to be greater than one and positive even if some individual labour-values are negative. Hence, according to them, positive surplus value is always guaranteed.

The problem with this argument is that its crucial assumption does not always hold. In fact, in Steedman’s example the value of the labour power is greater than one (to see that, just divide the variable capital by the total living labour) because of the non-separability problem discussed before. Thus, the argument put forward by ^{Trigg and Philp (2008}) that Kahn’s multiplier is positive - and that this implies positive surplus-value - for any system able to produce surplus even in the case of negative individual labour-values is correct only under the case where both final demand vectors fall inside the cone of feasible sets.

CONCLUDING REMARKS

The present work has proposed to give a clarification of the economic properties behind the “positive profits with negative surplus value” example provided by ^{Steedman (1975}). Criticisms of this example such as the lack of economic meaning of the presence of an inferior process seem quite misleading because the negative marxian aggregates are not a direct consequence of it^{11}.

Individual negative labour values are certainly a necessary condition for the paradox to happen but not sufficient. The central point seems to be the fact that the bundles that go to each class must: (i) fall outside the cone given by the square system and (ii) have a very different composition. So, the negative surplus value of the example is related to the fact that negative levels of activity would be “required” to produce only (with no overproduction) the capitalists’ bundle - something that is devoid of economic meaning. Hence the “phenomenon” of negative surplus value does not mean that in order to produce only workers’ consumption more labour would be required than in the case where there is also the capitalists’ consumption. It only means that the workers’ consumption bundle cannot be produced separately by this joint production system and, *thus, there would be some overproduction if there were no demand coming from the capitalist* (i.e.,, if the bundle were produced separately).

As the alternative provided by Morishima’s true values does not seem useful as being too “normative” we think that our improvement of the method originally proposed by ^{Akyüz (1983}) seems to be the simplest and at the same time least arbitrary way to get an economically meaningful and thus positive measure of surplus labour in viable pure joint production systems. The procedure has the advantage of using only data from the processes that are actually in use as opposed to the literature on the so called “fundamental Marxian theorem”.