Steve Keen on Marxist Economics, Together with
a Mini Essay on the Labor Theory of Value

A very one-sided review of Steve Keen’s Debunking Economics: The naked
emperor of the social sciences
, (NY: Zed/St. Martin’s Press, 2001), 335 pp.
(This review/essay focuses on the single chapter on Marxist economics,
and is also "one-sided" in focusing mostly on the labor theory of value.)

1. General comments.

Steve Keen’s book, Debunking Economics, is an excellent and important work. It is, for the most part, a sustained and devastating attack against the overwhelmingly dominant school of bourgeois economics, which is sometimes called the "neoclassical synthesis". For the uninitiated, that is the supply-and-demand school of thought.

Keen systematically demolishes all the major precepts of neoclassical economics and exposes a great many of its false assumptions, logical contradictions, and downright stupid dogmas. The supply-and-demand dogma itself is shown to be totally fallacious from both the demand and supply angles, as is the famous, but ridiculous, claim (deriving from Adam Smith) that every individual pursuing his or her own individual self-interest will somehow, via "the invisible hand", result in the greatest happiness for society as a whole. I was especially appreciative of Keen’s slaps against Jeremy Bentham in this area.

I readily admit I came to the book totally biased against neoclassical economics (and bourgeois economics in general). I haven’t taken bourgeois economics seriously since I suffered through a course using Samuelson’s famous (notorious?) textbook 40 years ago—and found it totally irrelevant for understanding the present screwed-up economic society, and also completely unresponsive to my already vaguely socialist interests in a totally different sort of society. To me neoclassical economics has about as much relevance to the real world, or to anything I am interested in, as does astrology.

A great many people, regardless of their political perspective, already know that "economics" and economists are pretty much a joke. In fact there are many actual jokes and humorous comments about them, though perhaps not quite as many as those about lawyers and politicians. Just a few short ones: "If all economists were laid end to end, they would not reach a conclusion." (George Bernard Shaw); "Nine out of ten economic laws are economic laws only till they are found out." (Robert Lynd); "To economists, the real world is often a special case." (Unknown).

In the issue of The Economist which just came today (dated Aug. 23, 2003) a reporter comments about the upcoming exclusive economic symposium which the Federal Reserve Bank holds in Jackson Hole, Wyoming, every year:

In the evening after dinner it has become a tradition for a select group of economists to meet for drinks on the terrace in the shadow of the towering Grand Tetons.

Since these include some of America’s top economists, your correspondent has made a habit of taking a poll each year. At the peak of the stockmarket boom they were asked, "is this a bubble?" "No", they voted. In August 2001 they were asked, "will America’s economy dip into recession?" "No chance", they replied. The economy turned out already to be in recession. Last year’s question was, "will the federal funds rate (then 1.75%) fall to 1% or less within the next year?" Almost everybody said "no" (a senior Fed official wisely declined to vote)….

Given this remarkable record, it would be remiss not to poll our elite group once again this year—and then bet on the exact opposite. We invite readers to suggest a question.

However, I know that despite all such jokes and sentiments and exposures, many people (including those at The Economist) do still imagine that there must be some considerable validity to (bourgeois) economics—despite its laughable track record when it comes to predicting and attempting to direct the capitalist economies of the world. For anyone who still thinks that neoclassical economics has any value whatsoever, I strongly recommend that you read Debunking Economics. I really don’t see how anybody who reads and understands Keen’s exposition can ever again take any aspect of establishment economics seriously.

But that does raise a small point. Economics is not a trivially easy subject, and neoclassical economics is especially abstruse. Keen does a pretty good job of making it intelligible to ordinary people, and manages to present all the important ideas without any use of mathematics beyond simple arithmetic. Even so, as Keen himself warns, "this book will still require significant intellectual exertion by the reader".

But if you are willing to put in at least this minimal amount of effort, you will come away really understanding why contemporary establishment economics is both totally erroneous and totally worthless.

And there are some nice little nuggets along the way. The very best of them, which comes after Keen debunks the neoclassical dogma that unemployment is caused by workers’ wages staying too high, is: "Neoclassical economics can be summed up, as Galbraith once remarked, in the twin propositions that the poor don’t work hard enough because they’re paid too much, and the rich don’t work hard enough because they’re not paid enough." (p. 76) Talk about a bourgeois economic theory!

2. A bourgeois critique of a particular bourgeois school of economics.

Despite my praise so far for Keen’s book, I have to say that it is a bourgeois book itself. It is a critique of one particular bourgeois school of economics from within a general bourgeois perspective. Even though it is a bourgeois book, it is still a good and important book just because it does such a good job in debunking this dominant bourgeois economic theory.

And I say that Debunking Economics is a bourgeois book not primarily because of Keen’s penultimate chapter which purports to also debunk "Marxian economics". I say it mostly because he himself argues from a perspective that assumes the permanence (or at least quasi-permanence) of capitalism into the indefinite future, and because he sees this not only as a given but as something for which there is simply no alternative. For him the possibility of socialism seems not even to exist.

There are many other indications of his essentially bourgeois outlook. When it comes to discussing economic crises, for example, what does he focus on? The stock market! As if the stock market was the most important concern even within the financial aspects of a capitalist economic crisis, let alone for such a crisis considered in all its major aspects. Moreover, most of that chapter doesn’t even talk about the causes of stock market crashes, but rather the various theories about how to predict the immediate trends of stock prices and thus make a fortune.

Despite his criticisms of neoclassical economics for viewing economic problems as being due entirely to "exogenous" shocks, how different, really, is Keen’s own perception here? He may indeed recognize that imbalances, to one degree or another, always exist within the economy, and even that chaos theory suggests that small internal causes may result in occasional large swings. But for him these are still accidental phenomena. In no way does he begin to understand that the real contradictions that are inherent in capitalism must necessarily result in economic crises—not because of accidents and aberrations and random chance events, but because of the very basic day-to-day workings of the capitalist mode of production. The accidents and chance events are not the basic causes of crises, but only the triggers.

While Keen’s book is written from an overall bourgeois perspective, I do have to complement him on both his acceptance and presentation of Marx’s critique of "Say’s Law". (This "law", briefly, is the notion that capitalist production automatically creates its own markets, and that there can thus be no gluts or overproduction unless there is some "external" shock to the system.) Keen rejects "Say’s Law" because he is a follower of Keynes, and specifically a "post-Keynesian". Paul Sweezy once went so far as to claim that "The gist of this Keynesian criticism [of orthodox economics] can be summed up simply as a flat rejection of what has come to be known as Say’s Law of Markets."[1] Whether or not that is completely true, it is a fact that the rejection of "Say’s Law" is central to Keynes’ viewpoint, and that Marx got there first and far more profoundly (as Keen recognizes). Yet strangely enough there is a strain within Marxist economics (going back to Lenin’s early economic writings) that accepts the validity of "Say’s Law"! So, bourgeois book or not, this is one of several areas where what Keen says is more correct and more sophisticated than what certain Marxists have said at times!

3. Marxist economics and the labor theory of value (LTV).

Of course it is in the one chapter on Marxist economics where many of Keen’s bourgeois biases and assumptions do come to the fore. One such immediate unfounded assumption is just that Marxist political economy can be refuted simply by showing that Marx’s version of the labor theory of value is untenable. It just isn’t so.

There’s a whole lot to Marxist political economy; many more aspects or principles than the two that Keen debunks (one version of the labor theory of value, and Marx’s "law" of the tendency of the rate of profit to fall over time due to the increase in the organic composition of capital). Even if we grant (as I more or less do) that these two aspects of Marx’s own theory are untenable, most of Marx’s edifice still remains, as does most of what subsequent Marxist economists have added to it.

No living scientific theory can be absolutely identified with every precise aspect of it as presented by its originators. Some things in Newton, Darwin, Einstein and the original theories of virtually every other great scientist have also been shown to be incorrect. But much of their thinking has been proven sound and to be of lasting value. We do not demand of any scientific theory that it be presented complete and perfect right at the beginning (or, indeed, that it ever be developed to the point where it is perfect and there are no more loose ends). On the contrary, two of the marks of a genuine scientific theory are 1) that it is continually modified, refined, and extended as we learn more, and 2) that it frankly recognizes the loose ends and anomalies that remain and sets to work on trying to resolve them.

It may well be true that the labor theory of value, as Marx presented it, is essential to Marx’s economics. But it is not essential to Marxist economics—that scientific theory which derives originally and mostly from Marx, but which has also been modified and extended (like any scientific theory) by many other people since Marx’s day.

Curiously, Keen argues almost the opposite of this, that while Marx’s labor theory of value (or LTV) is not essential to Marx’s own theory (or to one tentative version of it, anyway, that Marx hinted at), it must somehow, in practice, be essential to any "Marxist" economics that follows him. Keen only counts as Marxist economics those theories which are based on Marx’s original LTV. He says "I am as pessimistic about the chances of this ‘new, improved Marxism’ [without the LTV] being adopted by today’s Marxists as I am about the chances of neoclassical economists abandoning the concept of equilibrium." (p. 299) But there are actually a fair number of economists in the Marxist tradition who do reject the LTV (at least as precisely put forward by Marx, and usually altogether), and which Keen totally ignores.[2]

The fact is that Marxist economists, and others more loosely in the Marxist tradition, tend to be an enormously diverse and fractious bunch. They are not at all as homogeneous and invariably hidebound as Keen implies. And while many of them, it is true, tend to be somewhat doctrinaire or dogmatic, this is not true of all of them, and probably not even true of a majority of them in recent years. (In fact a good argument could be made that too many of them these days are too ready to part ways with Marx when there are not sound reasons for doing so!) To characterize and then dismiss all of Marxist economics because of the dogmatic characteristics of just some of them is not only unfair, it is downright unreasonable. In my opinion it shows an anti-Marxist bias on Keen’s part, despite some of the positive things he says about Marx himself.

4. The untenability of Marx’s labor theory of value (LTV).

Dropping the LTV (i.e., Marx’s version of it) actually serves to strengthen Marxist economics as a whole, and its theory of capitalist economic crises in particular. But I do recognize that many Marxists, especially those busily engaged in the revolutionary movement (rather than academic economists in the Marxist tradition), tend to blindly cling to the LTV despite all the logical and conceptual problems with it, which they are often completely unaware of. Much of the rest of this review/essay was written with folks like this in mind.

While I don’t dispute the essential validity of Ian Steedman’s Sraffa-like critique of Marx’s LTV which Keen presents, that is not what originally convinced me many years ago that Marx’s version of the LTV has major problems with it. Instead, let me present what I call the "android argument" against Marx’s LTV, which Keen doesn’t mention, but which I think is far simpler to understand, and hence far more compelling.[3] This is a sort of "thought experiment" of the kind that Galileo and Einstein were fond of. Galileo, for example, satisfied himself that (ignoring air friction) heavier bodies do not fall faster than lighter ones by simply considering that, if two weights were falling side-by-side, it would be ridiculous to suppose they would suddenly begin to fall faster if they got hooked together with a string and became a "single object".

Marx’s version of the LTV claims that only human labor can create surplus value (i.e., value beyond that which is used up in the production process), and specifically that machinery, no matter how complicated and sophisticated it is, cannot do this. (Thus this LTV denies that a machine could contribute more value collectively to all the commodities produced using it, than the amount of labor that went into producing the machine and its components in the first place.)

But the big question then is: Why can human labor create surplus value, and why can’t machines do so? After all, sometimes we can actually replace human beings in the labor process with machines! The one is sometimes substitutable for the other. Just what is so special about direct human labor when it comes to creating surplus value? And there is no good answer to this question.

But suppose we try to argue that this special property of human labor is due to some special characteristics of human beings themselves—say their intelligence, ingenuity, creativity, and so forth. The embarrassing question then arises, "What if these attributes turn out not to be unique to human beings after all?" Then we would be forced to agree that these other entities, which shared—to some degree—the "special characteristics" of human beings, could also contribute to surplus value in a system of capitalist production.

Specifically, let us first suppose that humans someday build an android, a mechanical "man", who can do everything a human being can do, maybe many things even better than a human can do. This android has a computer brain that is capable of memory, learning, thought, intelligence, ingenuity, creativity, and all the other mental capabilities that human beings have.

Now if human beings are granted the ability to generate surplus value based on their "special qualities and capabilities", there would be no reason to say that this android—which possesses those same qualities and can do anything a human being can do—should be denied this same ability to generate surplus value.

There is one quibble that is possible here. The generation of surplus value is only possible if the value of the commodities produced exceeds the value equivalent of the raw materials, overhead, machinery and labor power that is used up in the production process. But the value of the "labor power" of the android might be very different than the value of the labor power of a human worker. Let us suppose for the moment that the value of the labor power of the android works out to be identical to the value of the labor power of the human. (We might assume that this value equivalent for "android labor power" results from the labor time necessary to produce the android to begin with and the labor time required to keep it in working order, averaged out over the expected working life of the android.)

Let us now extend the thought experiment in a couple different ways. First, let’s suppose that the capitalist factories have constructed not just one such android, but hundreds of millions of them. Let’s suppose that all around the world all human workers are fired from their jobs (including at the android factories), and that only androids are "hired" and do all the work from then on. Let’s suppose the androids replace human beings because they can be forced to work much longer hours. (Yes, I know this is a horrible new kind of extreme wage slavery—or maybe even outright chattel slavery if, contrary to what we are supposing here, there are no android wages—since these androids should properly be considered sentient beings with the same "human" rights that natural human beings now theoretically have. But we may reasonably assume the capitalists will not be very concerned about these "fine points".)

In this situation all future capitalist production would still generate surplus value—and thus profits for the owners of the factories. (I assume there would still be a (partial) market for the factories’ output, since the androids would be paid wages, though the unemployed human workers would probably be left to starve.) Thus not only can one android conceivably replace one human being and generate the same (or greater!) surplus value through his/her/its labor than the human being does, but human beings can conceivably be completely eliminated from the capitalist production process by millions of such androids, without stopping the continued generation of surplus value.

So far all we have established in this thought experiment is that actual human beings are not necessary for the generation of surplus value after all; that entities which amount to artificial equivalents of human beings can do the job every bit as well. But now let’s extend our original android thought experiment in a different way.

Let us suppose that the first androids constructed are not really 100% equivalent to human beings. Suppose, first, that they are almost exactly the equivalent of human beings except that due to complex requirements of their computer brains (requiring long database reorgs, let’s say) they can only work six hours a day. During those six hours they can do anything a natural human being can do, but then they have to "sleep" for 18 hours. If such an android recovered its own labor power cost for the capitalist after 4 hours of labor, then for the next 2 hours each day it would be impossible to deny that it was producing surplus value—even if it couldn’t continue as long as a human being might be able to do.

Or suppose the first androids are only very close to other human capabilities, but just slightly deficient. Suppose, for the sake of argument, that IQ ratings are really meaningful (which they aren’t), and that these first androids have IQs of only 80. Just as we wouldn’t deny that a human being who was not very bright could nevertheless generate surplus value in many kinds of labor, we also could not reasonably deny that an android not as smart as the average human being would nevertheless be capable of generating surplus value in certain kinds of labor.

And if androids that are just a tiny bit less capable than humans can still generate surplus value, how about ones that are a bit less capable still, and then some that are considerably less capable? We can easily imagine a continuum of machines/androids from the totally unintelligent at one end to those which match or exceed human capabilities at the other end, and with virtually imperceptible degrees of difference between them.

This is a bit like the ancient Greek sorites paradox—is a man bald if you pluck a single hair from his head? Sure there is a qualitative difference between the extremes, but there is no particular line where you can say that on this side of the line all the men are bald and on that side they are not bald. All "human" attributes are like that, including every form of intelligence and creativity. They all come in degrees.

Well, you probably see where all this leads. Once we assume that it is the possession of certain characteristics and capabilities by human beings that allow them to be able to generate surplus value, then we are forced to admit that things or entities which possess those same characteristics—even to a markedly lesser degree—must also be capable of generating surplus value (though not perhaps to the same degree that human beings are capable of).

And the fact is that not only future androids, but also many of our sophisticated present-day machines already do possess qualities that were once assumed to be the exclusive property of human beings. There are already tremendous and varied aspects of intelligence, for example, in many different sorts of computer programs, and in computer controlled machine tools and other production equipment. Moreover, it is probably already true that a majority of human labor now being done could already be done by currently possible sophisticated machines instead. (The reason that this switchover is proceeding relatively slowly (on a year-to-year basis) is simply that humans are still much cheaper for most kinds of labor than employing the newest ultra-sophisticated and sometimes ultra-expensive machinery. But this too is changing rapidly in historical terms.)

We started out by asking just what is so special about human labor that supposedly makes it alone capable of generating surplus value in the capitalist production process. We tried out the explanation that it might have something to do with some special set of characteristics of human beings, characteristics that other entities don’t—or even cannot—share. The "android argument" and its extension to present-day machinery led us to admit that that ploy doesn’t really work.

I presented a very brief version of the above argument to a friend yesterday, but with my usual verbal ineptness. He objected, "But we don’t have any androids yet, and we may never have them." That was probably a valid objection to the argument as I presented it to him, but it is not a valid objection to the above thought experiment. Similarly, it would not be a valid objection to Galileo’s thought experiment to say "But you couldn’t tie the two weights together while they are falling because they are moving too fast!" Someone who said that to Galileo would be missing the import of the hypothetical experiment. It doesn’t matter if tying the two weights together while they are falling is a practical possibility or not! All that matters is what we would expect to happen if, somehow, they were tied together in that circumstance.

In the same way, it doesn’t really matter here if humanity ever actually creates any androids! What we are trying to think about is what we would say if such androids were constructed. And what we would clearly have to say is that if there were an "artificial man", an android, then that being would be capable of wage labor and of generating surplus value in exactly the same way that natural human beings are. And, in the second part of the thought experiment, to recognize that machines today (the more sophisticated computer-controlled machines, at the very least) already perform the equivalent of the labor of artificial human beings, and have their capabilities, at least to a considerable and ever-growing degree.

There is no set of special characteristics that only human beings alone can possibly share, nor is there even any such single characteristic that non-human entities (animals and machines) don’t already share at least to some limited degree. (And in the case of machines, to a very rapidly growing degree.)

Thus, philosophically, there is no tenable basis for the labor theory of value as put forth by Marx (and as is still defended by many—but by no means all—Marxists). This is no doubt something hard to accept for those who read their Marx years ago, and have long accepted Marx’s labor theory of value. But it is a test of the genuineness of our self-proclaimed scientific approach to politics and social theory how we deal with such revelations.

5. An invalid generalization of the LTV.

Suppose someone were to say, "OK, Marx’s original LTV is untenable, but there is a generalization of his idea which does not suffer from the same defect. In this generalized theory it is not human labor specifically which generates surplus value, but intelligence or creativity from any source, which alone can create surplus value. Thus machines can contribute to surplus value to the extent that they are intelligent or creative."

Note that this theory is in a sense intermediate between Marx’s LTV and a more general, Sraffaian-style, theory of the "production of commodities by means of commodities".

Would such a theory hold up? No, it wouldn’t. First of all there is the major (and probably insurmountable) problem of how one might independently measure the intelligence or creativity of machines. (It is, after all, something of an intractable measurement problem even for human beings themselves, partly because there are so many different kinds of things we call intelligence.) But there is a much deeper problem here than just that.

The fact is that there are indefinitely many specific things a human being can do while creating surplus value. It is by no means just a matter of "intelligence" or "creativity"! It depends upon the specific work he or she is performing. Often human labor which contributes to surplus value is still much more a matter of physical brawn than it is intelligence. Sometimes the intelligence exercised in human labor is very minimal, to the point that some assembly-line workers can get into day-dreaming trances while they perform their work like automatons, i.e., as if they were machines!

So even a weakly generalized version of the LTV which says that surplus value is due to the application of creativity or intelligence, by either human or machine, to the capitalist production process is quite inadequate. Moreover, any partial generalization of the LTV based on any specific list of human/machine attributes is still on the wrong track.

The real issue is not intelligence, or creativity, or any other characteristic or capacity of human beings—except for just the one truly relevant, abstract characteristic. And that abstract characteristic is the ability of human beings to create, through their labor (for whatever reasons!), more value than the value required to raise, train, and maintain their human labor power. Anything employed in a production process that contributes more to the value of the commodities produced than it uses up, thereby contributes to surplus value. That is what surplus value is, and that is what we have to look at directly, and not focus on any specific activity, of the endless array of possible activities, which the worker might be engaged in while creating surplus value.

Consider a simple tool, such as a claw hammer, for example, with which a worker constructs wooden freight boxes. A hammer possesses no intelligence, no creativity, and seems to share no other qualities that might be considered distinctively human. But a worker using such a hammer will be able to create much more surplus value for the capitalist than another equally skilled worker armed only with a rock. And the difference in the extra surplus value generated by the worker with the hammer will surely greatly exceed the value incorporated into the hammer itself, after very little time at all.

It is not only sophisticated semi-intelligent machines that can contribute to surplus value. On the contrary, even the most simple tool may well do so (assuming it is itself not invariably "used up" (destroyed) in the production of each new commodity, and so forth).

Not every commodity used in the production process can contribute to surplus value, however! The raw materials that go into the commodity being produced are used up in the process, and—as Marx said—only contribute as much value to the new commodity as is already incorporated into them. Raw materials, despite being commodities, do not contribute to surplus value.

Thus Sraffa’s theory, or at least the name for it, "the production of commodities by means of commodities", is a little too general and abstract when it comes to determining all the real sources of surplus value in the capitalist production process.

6. A valid generalization of the LTV.

But now for some major counterpoint with regard to the labor theory of value. I have agreed that not only direct labor, but also some commodities, and tools and machinery in particular, can contribute to surplus value in a system of capitalist production. But where did all these tools and all this machinery themselves come from? They of course came from human labor applied to earlier tools, machinery and other commodities. All the use value produced today comes from a very long series of steps wherein human labor was applied to previous commodities going back to the time when there were no commodities, and human labor was applied directly to the natural materials of the world around us to make the first human-created items with use value, and then later the first true commodities (things produced for sale). In short, ultimately, everything that is produced is still (so far!) the result of only human labor applied to the natural materials of the world. Moreover, the constituent quantity of abstract socially necessary labor time remains the only yardstick that allows us to compare the quantity of value in commodities, and thus to explain the center of gravity of their exchange values (to the first approximation). So there is still some major validity to the labor theory of value after all!

Was that too fast? Let’s take it more slowly. True, both human labor and other already existing tools and machines were used to make virtually all our presently existing tools and machines. And earlier labor, tools and machinery made those tools and machines, in a chain of production that goes back a very long time (long before capitalism!). But Keen is easily shown to be wrong when he claims that "This process can go on indefinitely, with each step further reducing the commodity content. But no matter how far back you go, you can never eliminate this commodity residue. If you could, then there would be some commodities that can be created with absolutely no commodity inputs—or in other words, by magic." (p. 289) And yet we all know very well that there was a time before human beings made any commodities; there was a time before there were even human beings; and there was a time before the earth itself even existed. So this "go on forever" logic is obviously ridiculous.

The fact of the matter is that every commodity that exists today is ultimately the product of human labor acting on the natural material of the world around us—the rocks and minerals, the plants and animals, the air and water, and so forth. Marx himself always insisted that labor was not the sole source of wealth, despite the assertions of many people that this is what he claimed. In fact, he made a big issue about this in his "Critique of the Gotha Programme" (1875) where he starts by quoting a sentence from the new programme of the German Worker’s Party:

First Part of the Paragraph: "Labor is the source of all wealth and all culture."

Labor is not the source of all wealth. Nature is just as much the source of use values (and it is surely of such that material wealth consists!) as labor, which itself is only the manifestation of a force of nature, human labor power. The above phrase is to be found in all children's primers and is correct in so far as it is implied that labor is performed with the appurtenant subjects and instruments. But a socialist programme cannot allow such bourgeois phrases to pass over in silence the conditions that alone give them meaning. Only in so far as man from the beginning behaves towards nature, the primary source of all instruments and subjects of labor, as an owner, treats her as belonging to him, does his labor become the source of use values, therefore also of wealth. The bourgeoisie have very good grounds for falsely ascribing supernatural creative power to labor; since precisely from the fact that labor depends on nature it follows that the man who possesses no other property than his labor power must, in all conditions of society and culture, be the slave of other men who have made themselves the owners of the objective conditions of labor. He can work only with their permission, hence live only with their permission.[4]

So Keen is quite wrong in claiming that only "magic" could create any initial commodities; they were simply created by the application of human labor to the materials of the natural world around us. And, moreover, the earliest commodities were created before the development of the capitalist system of production. (Commodities, i.e., things produced for sale, predate capitalism. Individuals, even today, may sometimes use their own labor to create goods for sale, without there being any capitalist arrangement (exploitation) involved whatsoever.) Thus the first true commodities were not created by using the commodity "labor power", but rather by the use of human labor before it itself became a commodity (as it most often is today). Keen seems to be very confused about all these sorts of things.

Those with a bourgeois outlook find it very difficult to believe that capitalism is not timeless, that it has not always existed, and that it will not always exist into the endless future. And even most of those who might admit that maybe there was a time before capitalism itself (i.e., feudalism, slave society, or primitive communalism) still cannot bring themselves to further admit that there was a time before commodities were produced. They really believe that there must "always" have been commodities, if not all the rest of the elements of capitalism.

*       *       *

It is possible to mean many different things by "the labor theory of value". Here are a few examples:

  1. Human labor is the source of all wealth (all use value). As Marx argued in his "Critique of the Gotha Programme", this is—strictly speaking—false.
  2. Human labor applied to the materials of the natural world is the ultimate source of all wealth (use value). This proposition is correct. (At least so far; if we get to the point where we have "android labor" it may need to be generalized a bit!) Nothing that Keen or Steedman or Sraffa said refutes this in the least. The contributions to wealth from our machinery are only indirect (or delayed) contributions of earlier human labor.
  3. All surplus value in a capitalist system of production is due to the direct (immediate) contribution of human labor. Marx (most of the time) thought this was true, as still do many (not all!) Marxists. But there are insurmountable logical and "philosophical" (conceptual) problems to this proposition (as demonstrated by Steedman and others), and it must be abandoned.
  4. All surplus value in a capitalist system of production is due to the direct (immediate) and indirect (via machinery, etc.) contributions of human labor, since humans built the machines. This proposition is correct, and is not at all refuted by Steedman, et al.
  5. The values of any two different things (which are useful in their own quite specific ways), including any two different commodities, can be compared, but they can only rationally be compared if they have some universal characteristic in common. This universal characteristic can only be the abstract socially necessary labor time (either direct or indirect) which goes into the production of each thing or commodity. This version or aspect of the LTV also remains completely valid (when properly understood), despite Steedman, et al.

So what is the source of capitalist profit? It is still surplus value. But in a current production process it is not only the direct, immediate labor that generates surplus value, but also a certain amount of earlier labor which created the tools and machines in the first place. When human labor builds a machine, that machine can contribute more than the value of the human labor that went into making the machine, when it comes to creating value in the form of all the commodities produced by that machine. (In other words, as we have been saying, machines can also contribute to surplus value.) But the machine itself was still produced, ultimately, by human labor (from natural materials), and so whatever extra amount the machine contributes to surplus value, that too is ultimately due to (past) human labor.

Things are a little tricky at this point, so let me try to clarify matters a bit further. I am claiming both that:

  1. A machine can contribute to surplus value in the commodities that the machine is used to create. That is, more value can come out of a machine than need go into it (when it is made and also in the form of maintenance). This is precisely the same as the case with human labor: human labor can create more value than is required to raise, train and maintain the human laborer.
  2. But despite this, all the surplus value in the commodities produced with the machine can still be considered as having its source in human labor, and specifically in a combination of "present labor" (the direct labor that goes into creating the final commodities by using the machine) and a portion of the "past labor" (the labor that went into creating the machine).

But how can both these things be true? Only because machines allow us to reuse past human labor over and over again! (That is, the labor that went into making the machine.) That is really what tools and machines are for, when considered from the most abstract point of view. (It is somewhat similar to a subroutine in computer programming that—once written—can be called (used) over and over again, except that unlike physical machines, program subroutines never wear out.)[5]

Conceptually, when "we" (speaking of humanity as a whole) use a machine in a production process we are in effect going about things in a very roundabout way. We first make the machine, then we use the machine to make what we really want to produce. If we were only going to make one item with the machine this would probably be foolish. But since we can reuse the machine to make many items, it soon pays off to take such a roundabout path. It seems quite strange, but technical progress in production generally means taking ever more circuitous paths towards our final production goals.[6]

At one point Keen quotes a passage from volume one of Capital in which Marx denies that machines can contribute to surplus value:

The maximum loss of value that they [machines] can suffer in the process, is plainly limited by the amount of the original value with which they came into the process, or in other words, by the labour-time necessary for their production. Therefore, the means of production can never add more value to the product than they themselves possess independently of the process in which they assist. However useful a given kind of raw material, or a machine, or any means of production may be, though it may cost £150, or, say 500 days' labour, yet it cannot, under any circumstances, add to the value of the product more than £150.[7]

So what is wrong with Marx’s reasoning here? It is simply this: He is right in claiming that no more than the value embodied in the machine can be transferred to any single commodity produced with the machine, but he forgets here that the machine can be used over and over again to transfer a certain amount of value to each of a great number of output commodities. No single commodity produced with the machine can have more than the value embodied in the machine transferred to it, but all the commodities produced taken together can collectively acquire much more value than the value of the labor power that went into making the machine. This is the main reason why it is worthwhile to make machines in the first place! If it weren't for this fact, building machines would typically be a wasteful, needlessly roundabout approach to our actual production goal, which could usually be tackled much more directly.

It is true that this generalized LTV makes it more difficult to measure surplus value, and also introduces various other complications. But there is no fundamental logical or philosophical problem remaining here.

7. Do machines and human beings contribute to surplus value in the same way?

That’s an interesting little question to consider, though perhaps the basic answer is obvious. I’ll first quickly state my conclusion: Yes, they do!

Machines (and also simpler tools) may contribute to surplus value just because they can be used over and over again. The same goes for human beings. There is nothing unique to human beings in this regard, let alone anything "mysterious" about our capacity to create new value. Unfortunately, Marx, by failing to explain just why human labor "alone" can create surplus value, inadvertently gave the impression that this is mysterious. The value (labor time) that goes into raising, training, and maintaining a human worker produces an entity that can be used over and over in the production process. That simple characteristic is the essence of the matter here, and there is no mystery about it at all.

Humans are much more versatile than even the most sophisticated present-day machines. But there are many different types of machines which can each do one of the many specific things a human worker can do. In many common production processes, where only the ability to do some one specific thing matters, there is no effective difference between such a machine and a human worker. They both do the job, and with the same results. (Though one typically does it cheaper, which is why that is the one that is generally "employed" in that case.)

At one time all machines were relatively simple, and the tremendous gulf between them and human beings was obvious. Machines were appropriately contrasted to human beings as something entirely different, sharing very few characteristics, and even then only to a small degree. But things are rapidly changing. One qualitative leap occurred with the advent of computers. Additional leaps occurred with the introduction of numerically controlled machine tools and programmable assembly line robots. But much more striking advances are in store over the next few decades. We are in the early stages of the era of the development of certain special machines to the point where they can match, and then exceed, most or all human capabilities. Whether or not actual androids are eventually constructed (that both look and act like human beings)—and I think they will be—the fact is that the hard line between human and machine is fast blurring and disappearing.

Marx was wrong even about the ability of simple tools and machines to create surplus value. But maybe it will be easier for people to agree to this as some machines themselves become more human. (That’s part of the point of the "android" thought experiment.)

8. The "transformation problem".

One of the important things to recognize about this generalization of the LTV (which recognizes that some surplus value is also generated by—or via—machines) is that the "transformation problem" becomes much less of an issue.

Recall that the problem mostly arose because some industries use more machinery than do others. Thus if the rate of exploitation of workers, s/v, is assumed to be the same across all industries (as Marx did assume) then the rate of profit, s / (c + v), would have to be initially much lower in industries that used a lot of machinery (because the denominator is larger while the numerator stays the same). But if the capitalists found that they made less profit in one group of industries rather than the other, they would tend to shift their investments out of those low-profit industries and into the higher-profit industries. This would cause prices for products in the industries using greater amounts of machinery to rise, and prices for products in the other industries to fall. This in turn should lead to prices systematically diverging from values—either higher or lower—depending on how the amount of machinery used in that particular industry compared with the average industry. Consequently there needed to be a means of systematically deriving actual prices from labor-equivalent values, or put the other way around, a method of "transforming" values into prices. The trouble was, with Marx’s version of the LTV, there was no logically consistent way of doing this. (This is what Steedman, using Sraffa’s general approach, demonstrated.)

However, under the generalized LTV that I am defending, an industry with a greater amount of machinery would not have a lower rate of profit in the first place. These industries will merely be indirectly applying a greater proportion of past labor (in the form of machinery) and a smaller proportion of current labor. But since surplus value derives equally from both types of labor there is no problem here. An industry with more machinery than the average simply derives a greater percentage of its surplus value from past labor. But the rate of exploitation should still be same as any other industry, as should the rate of profit. (Of course we are ignoring monopolistic distortions here, and so forth.)

Naturally we do still have to recognize that prices will often diverge somewhat from values to some degree for a number of different reasons (including production miscalculations, unforeseen developments, and consumer fads). But there is no longer a systematic problem requiring a systematic solution.

Or is there?! It seems that we might now have a bit of a problem in the other direction! Because machines can be used over and over the capitalist can in effect readily reuse the labor that went into making those machines. For a fixed outlay for the machine, they receive more than the usual amount of labor power, and therefore more than the usual amount of surplus value. Thus it seems that there is now more surplus being produced with a given outlay of capital, and therefore a higher rate of profit, in industries which have more machinery than the average. This is just the opposite of the problem with Marx’s version of the LTV! Note also that according to Marx’s LTV, the capitalists should be biased against machinery since it serves to lower the long-term rate of profit (even if it gives the first capitalists to utilize new machinery an initial advantage over the others). Whereas, according to this generalized LTV, capitalists should be biased in favor of machinery for reasons both of initial advantage and long-term profit. I think that empirically the evidence pretty clearly shows that capitalists, in general, are quite biased in favor of machinery.

I’m still somewhat unclear about the validity and significance of this suggestion that profits should tend to be higher in industries with a greater than average amount of machinery, and I need to think it over some more. But at the moment it does seem to me that this generalized LTV also leads to systematic (inverse) divergences of prices from values in industries with either greater or lesser amounts of machinery than the average industry, and therefore to its own "transformation problem". But it also seems to me that this transformation problem is easy to solve and that there would not be the sorts of logical difficulties that arose for Marx’s LTV—because those problems really did arise from his invalid refusal to admit that machines can contribute to surplus value.

Even though I think that this can all be resolved rather easily, I would admit for now that it remains something of a loose end for the theory I am defending.

9. An example of how this generalized LTV works.

Suppose there is a capitalist factory that makes one commodity, "widgets". To make these widgets the factory owners buy raw materials (say steel and plastic) from other factories, and hire human labor (i.e., purchase human labor power). And they also buy one more thing, "widget machines", from yet another factory. (We’ll ignore all forms of overhead to keep things simple.)

Now suppose the following additional things:

So far we have accounted for 0.1 + 0.2 + 0.3 = 0.6 (or 60%) of the value of each output widget. The rest of its value (i.e., 1.0 – 0.6 = 0.4) must be surplus value, which we now assume comes from both direct human labor and the machine. (I won’t call it "machine labor", since it is not yet from a sentient android!). How then do we apportion this surplus value between the two sources?

One option is to simply say we have no basis here for deciding this. In this case, we just say that the surplus value in each widget is 0.4 (or 40%), and that this surplus value comes from a combination of direct labor and the older labor that went into the production of the widget machine. This is the simple way out. Doing things this way we have:

Note that the profit rate here comes out to be exactly the same as it would under Marx’s version of the LTV. However the rate of surplus value must now change since variable capital is a larger proportion of the total. So in this case the rate of surplus value, s/v, would be 0.4 / 0.5 = 80%. For Marx it would have been s/v = 0.4 / 0.2 = 200%. I don’t see that this difference matters very much, despite the large numeric change. In fact, as I suggested in the last section, this actually helps resolve the central difficulty in the "transformation problem".

Philosophically, you could argue that attempting to divide up the surplus value between direct labor and old (indirect) labor makes no more sense than attempting to divide it up between the 14 specific tasks (let us suppose) that a worker directly performs in making each widget. Omitting any one of these steps would mean that the widget could not be completed, i.e., could not be produced. All are essential, in that sense, and none is more essential than any of the others. You could, however, apportion the value based on the various lengths of times that each of the 14 steps takes.

But suppose we know that workers can produce 3 widgets/day if they do not have widget machines (and remembering that they can produce 10 widgets/day with the machines). Then 30% of the widgets produced per worker per day could be attributed to the human workers, and 70% could be attributed to the machines. And therefore 70% of the surplus value could be attributed to the machines too. In this case it would mean that .7 × 0.4, or 0.28, is the surplus value which we can attribute to the machine.

However this sort of calculation won’t really work. Consider, for example, the situation where no widgets at all (or maybe only 1 per year) could be made by a worker without the use of the widget machine. In this case, if we used the above logic, we would have to attribute all (or virtually all) of the surplus value to the machines! And yet we know that without direct human labor added, the widget machines are totally useless. It just doesn’t seem reasonable to attribute all of the surplus value to machines except arguably in the case where machines truly do all the work, and humans are no longer needed at all. But this consideration now discredits our previous case where we decided that the contributions to surplus value from the machines should be pegged at 70%.

Fortunately there is a more rational method here. What we can do is the same thing that was done with the 14 direct labor steps above. That is, compare the labor time necessary to produce the machine itself to the direct time necessary to produce the widgets by using the machine. Let us suppose that the value of the labor (i.e. the total labor time expended) that went into making the widget machine (and its components!) was equivalent in value to 200 widgets. So the total labor time required to make 1 widget by using a widget machine should include both the direct labor time, 0.2, and the indirect labor time, 200.0, for a total labor time equivalent to 200.2 widgets. But the total labor time per widget necessary to make two widgets is only 0.2 + (200.0 / 2) = 100.2, since we can freely reuse the labor time it took to build the widget machine (at least until it breaks down!). Let us assume that in the normal course of things the widget machine is used to the fullest, that is until it breaks down after making 1,000 widgets. Then the labor time contributed to each output widget through the widget machine is 200 / 1000 = 0.2, and the direct labor time per widget is still 0.2. In other words (in this specific case) 1/2 of the 0.4 amount of surplus value should be attributed to the machine (i.e., the past labor that went into making the machine), and 1/2 of it should be attributed to the direct human labor.

We could get into more complicated examples and additional considerations, but since this is supposed to just be a book review and not a treatise on the labor theory of value, I’ll leave it there for now.

10. Keen’s false conclusions from the rejection of Marx’s LTV.

Keen is correct to criticize both Marx and those followers of Marx who hold to Marx’s original version of the labor theory of value. (That is, the version that says that surplus value comes exclusively from direct human labor.) I don’t in any way object to his making those criticisms. But he is very wrong in some of the conclusions he draws from the rejection of Marx’s LTV.

First, as I already said, he is wrong to draw or imply the central conclusion that the untenability of Marx’s LTV means that Marxist economics has been refuted. Yes, Keen does say that if Marxists can just reject the LTV then they might have something going for them. But he also treats Marxist economics as "debunked" just as he much more correctly does neoclassical economics. In his last chapter, on the alternative economic theories which might replace the debunked neoclassical theory, there is no mention that Marxist economics might be the way to go here. On the contrary, he mentions only a mishmash of half-baked bourgeois theories, all of which seem to have varying numbers of elements in common with the thoroughly debunked neoclassical economics.

Second, Keen says that the rejection of Marx’s LTV means that capitalism is not exploiting the working class after all. This is an utterly ridiculous conclusion. It is true that direct labor is not the only source of surplus; some of it does come from the machinery, the "capital", that the capitalist provides. But where did the capitalist get that capital, those machines and so forth? From the earlier exploitation of workers!

Just because the capitalist "owns" the machines, it does not follow that the capitalist is really contributing to the value of the output commodities. After all, the capitalist "owns" the labor power of his workers, too, since he bought it (i.e., is paying them wages). You could just as well argue that the workers contribute nothing to the value of the output commodities, that it is all due to the contribution of things owned by the capitalist! No doubt this notion might be amenable to the capitalists! But to everybody else it is obvious baloney. In reality everything that the capitalist "owns" came originally from exploiting labor (or else through other forms of crime, other forms of theft which were dominant at the time of primitive accumulation, but often continue today as well).

Third, Keen says that "Marx used the labour theory of value to argue that capitalism harboured an internal contradiction." (p. 271) Since this LTV is invalid, says Keen, then so is Marx’s conclusion that there is an internal contradiction which means "that capitalism must necessarily give way to socialism." There are many things wrong here. First of all, neither Marx nor any serious current in revolutionary Marxism since Marx’s day, has ever argued that capitalism must automatically collapse regardless of what anybody does.[8] On the contrary, revolutionary Marxists have always maintained that capitalism will be overthrown precisely because of the misery it brings about in the working class and among the masses, and consequently because of what this then leads the masses to consciously decide to do. In short, the implicit "automatic collapse" theory that Keen invokes has never been part of genuine Marxism, and certainly not since Lenin’s vociferous attacks on the notion. Next, the fact that Marx’s version of the LTV is untenable, does not mean that the more sophisticated, generalized version of the LTV I mentioned above (in section 6) is untenable. And this means that there is still the very same fundamental internal contradiction inherent in the capitalist mode of production itself, namely the contradiction between social production and private appropriation that leads inevitably to major economic crises.

But Marx talked not only about the contradictions which lead to capitalist economic crises, but also to the means by which these crises are overcome. (The central means is through the massive destruction of capital, of the means of production itself, which clears the ground for a renewed expansion.) So, once again, the fact that capitalism does have this internal contradiction does not mean its "automatic" collapse. But this internal contradiction does, nevertheless, mean that capitalism is incapable of fully meeting the needs of the people of the world (especially outside the imperialist centers), and is therefore always generating revolutionary movements against it. And this will continue until either capitalism is overthrown, or until humanity has been destroyed by capitalism. Unfortunately this last possibility must be taken ever more seriously.

11. The "law" of the tendency of the rate of profit to fall.

I won’t say much about this supposed law that Marx talks about in chapters XIII-XV of vol. 3 of Capital, except that I agree with Keen that it is invalid. However Keen fails to see that this "law" is by no means essential to Marx’s theory of economic crises.

In fact, to be terribly frank about it, Marx did not live to work out any single unified theory of capitalist economic crises. Instead, there seem to be three main theories in his writings which fit together very poorly:

The overproduction theory is really the only theory that Marx needed, and he actually weakened his overall crisis theory by adding in the other two. There is some limited truth to the anarchy theory; that is, it is a small secondary factor which can contribute to and complicate economic crises. But the falling rate of profit theory is completely wrong. (Profits do fall, of course, in crises, but this is a result and not a cause of the crisis. They fall because markets collapse and goods which have been produced cannot be sold at full value. Sometimes they cannot be sold for any profit at all.)

So Keen is wrong to think or imply that just because Marx’s version of the LTV is untenable that his whole crisis theory is invalid as well.

12. Conclusion.

What are the marks of a really good book? There are many of course, but one of the primary ones is just that the book really makes you think. This of course means that what is a good book for one person might not be a good book for another. But in my case at least, Debunking Economics really did make me think about a lot of things. First of all, it made me think a lot about the main topic, the neoclassical economic theory and its many fatal flaws. But much more importantly for me, Keen’s book really provoked me to start thinking more deeply about the foundations of Marxist economics, and the labor theory of value in particular. Although I already long-ago recognized that Marx couldn’t be right in saying that machines cannot contribute to surplus value, I hadn’t really sat down before and tried to think out the many implications of that simple recognition. I don’t say that this thinking out of all such important implications is now complete for me; far from it! But at least I have at last gotten underway on the project.

So I want to sincerely thank you, Steve Keen, for writing this book. It has been important for me!

—Scott H. (Sept. 2, 2003)
    (Edited to correct confusions of use value/exchange value, etc. on Dec. 4, 2003.)


[1] Paul Sweezy, in Science & Society (Fall, 1946); quoted in Maurice Dobb, Theories of Value and Distribution Since Adam Smith (Cambridge Univ. Press, 1977 (1973)), p. 216f.

[2] Some of these economists in the Marxist tradition who reject Marx's labor theory of value hang out on the Progressive Economists News List on the Internet. For info on subscribing to this news list, and an archive of past postings, go to:

[3] I don't claim to be the only person who has thought of this general sort of argument, which has roots going back at least to Hilferding, I believe.

[4] Marx, "Critique of the Gotha Programme", (Peking: Foreign Language Press, 1972).

[5] Actually, any particular physical manifestation of the subroutine (or complete program) can "wear out". For example the pattern of magnetic domains on a hard disk that encodes the subroutine will eventually degrade. But once written, a subroutine or entire program can be preserved (and thus reused) indefinitely, by simply copying it to new hard disks, etc.

[6] Somebody else said something like this before me, about progress coming from ever more circuitous production methods. I forget who it was.

[7] Marx, quoted in Keen, pp. 278-9. (I haven't looked up the passage myself, but it comes from comes from vol. 1 of Capital.)

[8] It is true that some individuals, especially those of dubious revolutionary credentials, have argued along these lines. Bebel, for instance, argued at the 1891 Congress of the SPD that "bourgeois society is working so vigorously towards its own destruction that we need only wait for the moment when we can pick up the power which has already dropped from its hands." [Quoted in Howard & King, A History of Marxian Economics, Vol. I, p. 72.]

— End —

Marxist Political Economy Home Page
Scott H.'s Home Page on MASSLINE.ORG