Defending the neoclassical position, whatever that is, is not really my main business here today. We are here to talk about Luigi Pasinetti's very interesting book (Pasinetti: 2007).

The General Theory of Employment, Interest and Money, Keynes's great work of 1936, is a difficult book because it contains several different themes. It is not always clear that the different themes are compatible with one another, so the book is very hard to summarize or to give to beginning students to read. I think Luigi Pasinetti's work shares some of that quality; there is more than one thing going on and it is not always clear – at least not always clear to me – exactly how the several themes fit together.

I am going to focus my discussion on one of those themes: ‘A Revolution in Economics to be Accomplished.’ Luigi Pasinetti argues that the Keynesian Revolution was left incomplete and that the Cambridge Group (and a part of the book consists of very affectionate and very nicely written portraits of members of the Cambridge Group) was trying to complete it, but did not succeed. Professor Pasinetti's own project is in fact to contribute to completing the Keynesian Revolution. I don't remember that he ever says very clearly, or even not very clearly, exactly what the Keynesian Revolution was and how it is incomplete, what the remaining gap is.

What arises from the pages of the book – and Will Baumol touched on this – is that the Keynesian Revolution was somehow opposed to more or less everything in orthodox economics, in what is today called neoclassical economics. Without further details, it is almost as if we were trying to agree to meet somewhere in Rome and I asked you where, and you said, not in the Piazza Colonna. Yes, I understand that, but it doesn't help me very much about where exactly to go.

I am going to take a very particular view. I am going to state what I think Keynes's General Theory accomplished. I am going to agree that it is incomplete, but I think that it is perhaps inevitably incomplete for a reason that I will describe. Third, I am going to maintain that the Cambridge Group was not really engaged in completing it, but that others were, and not in the direction that Professor Pasinetti prefers.

You should keep in mind all the time that I am exactly what Joan Robinson called a ‘bastard Keynesian.’ Joan Robinson intended that label as a devastating insult, to make the bastard Keynesians ashamed of their origins and of their actions. But I rather like the label and I adopt it cheerfully. I am sure that if I were a Freudian I would be a bastard Freudian; and if I were a Marxian I would be a bastard Marxian; and so I am indeed a bastard Keynesian.

In my view, Keynes freed macroeconomics from its dependence on market clearing, especially, but not only, in labor markets. In the older view, in the view that Keynes opposed, there were indeed what Pigou called lapses from full employment. They had to be explained mainly by temporary deviations of prices and wages from their proper values. Keynes, in opposing this, argued that those lapses were persistent and normal and could not be expected to cure themselves. Of course, Keynes argued in the General Theory much else besides, but that, it seems to me, was the main point of the Keynesian Revolution.

Keynes, by the way, thought that he had worked out that a modern economy could be in equilibrium with unemployment, meaning that there were no internal forces at all tending to move it out of that state. Most modern economists, including bastard Keynesians like me, think that Keynes did not quite make good on that assertion. He simply lacked the analytical tools. Long after the General Theory, several economists showed how particular not wholly implausible assumptions – but not wholly plausible either – could indeed in principle give rise to an under-employment equilibrium. These were mostly assumptions about expectations. When you set your mind to it, it is not hard to visualize situations in which, if there are almost universally pessimistic expectations about the immediate future, actions will be taken which indeed bring about the pessimistic state and thereby confirm the pessimistic expectations. But in the same situations, if for some reason universal expectations had been optimistic, they would have given rise to actions which would have brought about a better state and thus confirmed the favorable expectations. But those are rather special assumptions and I don't want to direct attention to them now.

Most of us bastard Keynesians think that this failure to provide a tight notion of equilibrium with unemployment is unimportant. The Keynesian revolution survives quite well if the restoring forces in an under-employment disequilibrium, say, are so weak and so slow that it would take unacceptably long for a satisfactory equilibrium to be restored; and if there are fiscal and monetary policies available to the state that would shorten the period and drastically reduce the social cost. For a bastard Keynesian like myself, the Keynesian Revolution would then have won the battle, and indeed that is what I think is the case.

For long enough periods of time, disturbances occur frequently enough that aggregate output is determined by aggregate demand with persistent excess supply of output. Other possibilities exist, of course, but that is the one that one wants to focus on. That is my idea of what the Keynesian revolution was: the direction of macro-economic thought to situations in which aggregate output, not necessarily in every market, is limited on the demand side, and thus there is persistent excess supply at the aggregate level. That battle, to my mind, was won, and I never had any hesitation about that.

Like any good intellectual revolution, the Keynesian one opened up large vistas of useful research on the behavior of consumer spending, on the behavior of business investment spending, on the mechanisms by which monetary policy and fiscal policy can affect aggregate demand, and therefore output, when output is limited by effective demand. There is room for research on the behavior of labor markets, and credit markets, on portfolio choice, on the role of expectations, on wage formation and price formation.

You will see from what I have said that the Keynesian Revolution will never be complete because behavior and institutions in our societies change. They evolve. And what are the correct answers, or at least reasonable answers, to those questions now or five years ago may be inadequate answers five years from now. I hope small armies of young economists will be engaged in trying to bring the Keynesian Revolution up to date. But – and this is where I differ very strongly from Professor Pasinetti – almost all of this work of extending and completing and embedding the Keynesian Revolution has been carried out by bastard Keynesians. In fact, I do not myself see that the Cambridge Group was carrying on Keynes's work in any genuine way. Similarly, I am inclined to question whether there is much of a connection between Keynesian ideas and classical economics, if by classical economics we mean economics free of demand-side considerations.

Professor Pasinetti's interest in classical economics seems to come from a belief that demand considerations are bound up with social institutions in a way that supply-side considerations are not. So the supply-side considerations can therefore be described as ‘natural.’ I am not really convinced by this distinction, neither its reality, nor its significance, because, on the production side, productive efficiency is already an institutional consideration. Robinson Crusoe's decisions are a mixture of technology and preferences. In fact, once we get away from the realistically uninteresting case of one primary factor and no technological substitutability, choice plays about as big a role on the production side as it does on the consumption side. The motives and information and rules of thumb of firms are just about as important as utility functions and other such things for consumers, so the productions side is not at all free of institutional influences. There's nothing natural about the classical model.

We take up here another question. In the latter part of the book Luigi Pasinetti seems to want to draw a related distinction along these lines: that if it is not the fundamental cause of the failure of orthodox economics, it is at least the symbol of the failure of orthodox economics that it is somehow the economics of an exchange economy with the quantities of goods produced already given. So it provides only an elaborate working out of the consequences of market exchanges, whereas the unfinished part of the Keynesian Revolution will somehow give proper primacy to the production side of the economy and be an economics of production.

I have to say that I really do not understand this distinction. I remember dimly from the 1960s that it was one of those oracular statements that Joan Robinson used to make which were intended to end discussion, not to begin it, but in fact I don't think it is accurate. When I was sitting in my office at MIT making notes for this discussion, I could not find my copy of Arrow and Hahn's book on competitive general equilibrium. (I had presumably lent it years ago to a student who is now teaching in California, or something of that sort.) But I did find my copy of Gerard Debreu's Theory of Value. When I opened it I discovered that in Debreu's book, the chapter on the theory of the producer comes before the chapter on the theory of the consumer. The notion of the production set is prior in his discussion to the study of markets.

Then I managed to put my finger on another old book from my younger days, Jacob Mosak's book on General Equilibrium Theory and International Trade, and there I found exemplified exactly what had always seemed to me to be the case: the prominence given in orthodox treatments of general equilibrium to the pure exchange case is a purely pedagogical device. The idea was that it is easier to solve, to demonstrate to the students, or to oneself for that matter, how equilibrium theory works out, if the supply of goods is already given. Then if you now add a production side to the economy, you have all that earlier work available to be used as a sort of lemma in producing a theory that involves the production side as well. When I opened Mosak's book I found that he had done exactly that and said that it was what he was doing. He treats the pure exchange case first, he says, because then it will be easier, once we have these results in our pockets, so to speak, to allow for the production side. He then goes on in the very next section of the book to elaborate his international equilibrium with production. So I really don't think that any such distinction characterizes the difference between orthodox economics, neoclassical economics and the new Keynesian economics to come.

It is interesting that, as I read Professor Pasinetti's book, I found that the most interesting and exciting pages in it are precisely the pages in which he is talking about his own work, not about the unfinished Keynesian Revolution as a drama of some kind. I think that Professor Pasinetti's own work on what he calls structural dynamics, on essentially how to extend the study of macroeconomic dynamics to incorporate the structural changes – some endogenous, perhaps most endogenous, but at least some still exogenous – in an analytical framework that is capable of looking at the evolution of capitalist economies. I am wholeheartedly in favor of that. And there Luigi Pasinetti has some important predecessors. Hollis Chenery, for instance, my old friend from graduate student days, was trying to do exactly that in his analysis of developing economies long ago, but without the elegance and without the analytical detail that Professor Pasinetti is able to bring to bear.

Another very distinguished and able economist, Leif Johansen, the Norwegian, wrote a Ph.D. thesis which was a multisectoral model of economic growth in which he too proposed to try to provide a framework in which structural changes can be embedded naturally and analytically. I am all in favor of this. I think that is exactly an open question. I think it has very little to do with the Keynesian Revolution, but has everything to do with improving our analysis of an evolving capitalist economy.

To give you a concrete example – and as a bastard Keynesian I am a great believer in concrete examples – I think that modern analyses of economic growth have paid too little attention to the distinctions between the production of goods and the production of services. It is very odd for me to be saying that, since I have spent a certain amount of time trying to explain to over-enthusiastic non-economists that goods and services are both examples of economic goods and that they obey many of the same laws, have many of the same incentives. But in fact there may very well be systematic empirical differences in the capital intensity of goods production and the capital intensity of production of services. And there may very well be differences in the pace and role of technological change in the service industries and in the goods industries. This is a case of structural dynamics and today Luigi is one of the few economists who is paying serious attention to such things. But it is not in any way related to the Keynesian Revolution.

Let me summarize the world as seen by a bastard Keynesian. There was a Keynesian Revolution. It was the Keynesian Revolution most of us know and love. It focused the attention of the profession on the macroeconomic importance of aggregate demand and argued that aggregate output often was limited by aggregate demand; that policy mechanisms in periods of excessive unemployment and idle capacity should be aimed at increasing aggregate demand. Those lessons are still with us. They are in yesterday's newspaper and I hope in tomorrow's newspaper in the European Union, in the United States, in the People's Republic of China, and everywhere. I think that was a true revolution. It was much opposed at the time. I think it remains and will always remain to be completed because we will never – I presume never – come to the end of the evolution of the forces that govern aggregate demand and the way labor markets and commodity markets and service markets translate those things into behavior that we observe from time to time.

Now I will say one word about how all this fits with neoclassical growth theory, with which I seem to be associated. I don't know how many people in this room have read that 1956 paper of mine which determined so much of my career, as I did not realize it would at the time I was writing it. If you have read it and if you have read it to the end, you will remember that after about two-thirds of the text, there is a section that begins with words something like the following. Everything so far has been the neoclassical side of the coin and now we have to talk about the Keynesian aspects of the problem, the fact that growth paths for capitalist economies are very rarely paths of continuous equilibrium at full employment. Indeed they are not. But there is a puzzle, and this is another puzzle that perhaps is an uncompleted part of the Keynesian Revolution. It relates to something else which is an emphasis that I endorse and have endorsed in the past. I think that the economics of the medium run is terribly important and not worked out for the following reason. While it is true that the growth paths of capitalist economies are never paths of continuous full employment, full utilization equilibrium, it is true as a matter of historical fact that it would be reasonable to describe those paths as fluctuations around a trend, with most of the time the fluctuations contained within fairly narrow bounds – a couple of percent, three or four percent on either side of the trend path, defining what Axel Leijonhufvud once called the corridor around that path.

Only rarely – the Great Depressions of the 1890s in the US, the 1930s in the world at large, and one doesn't want to guess about the next few years (one no longer has to guess) – are there very large divergences from that path. I always thought (and that is what I think you would read into that 1956 paper, if you read it through to the end with some sympathy) that the goal of neoclassical growth theory was simply to try to understand the forces that underlay the trend. It's not at all engaged in the problem of describing actual paths, because actual paths consist of successions of short runs. What is lacking, and where I think the medium run is of such fundamental importance, is how to knit together those two things. Keynes after all wrote in the General Theory as if the stock of capital were constant. He was explicit about that. Investment is taking place, but it is small relative to the existing stock of capital. So he would treat the stock of capital as exactly constant. In the long run obviously one cannot do that, since one of the obvious, undeniable, visible aspects of growth paths is that the stock of capital rises.

How does one make the analytical connection between the short run and the long run? In other words, what happens in the medium run when we can no longer treat the stock of capital as a constant, but we have somehow to carry on an analysis of aggregate demand and aggregate supply?

I have made a couple of attempts at that; other people have, too. I don't think that that problem is solved and I hope one can continue to try to solve it. I am afraid that the tendency amongst very modern macroeconomists (I cannot know what takes place in Italian universities, but I can tell you that if this discussion were taking place at the University of Minnesota in the US, the student body would be totally uncomprehending) is that they would regard every moment not only in the long run or in the medium run but in the short run as well, as an example of full macroeconomic equilibrium. This strikes me as madness. But I am just an old bastard Keynesian, after all.

The preceding remarks originated in a discussion, at the Accademia Nazionale dei Lincei, of Luigi Pasinetti's Keynes and the Cambridge Keynesians. Naturally, then, it consists mostly of comments on the work of the Cambridge Group and on the significance Pasinetti attaches to that work. The developments dedicated to structural dynamics are rather limited in the preceding exchange: an expression of strong support of the basic idea, a reference to earlier strands of work on multisector growth models and patterns of economic development, and the concrete suggestion that analyzing the sources and consequences of the well-documented shift from goods to services in modern economies is an excellent example of structural dynamics waiting to be undertaken. I should have seized that occasion to emphasize that the shift seems to be a response to the relatively high income elasticity of demand for services such as education, recreation, travel, food preparation and health care, and therefore a clear example of the need to think about demand side and supply side together. I would like to add a few more general remarks here.

I cannot imagine how anyone could be ‘against’ the goal of a structural dynamics. Multisector growth models and one-sector growth models are complements, not rivals. The main reason for pursuing one-sector and two-sector analysis is transparency. The role of certain fundamental principles, like the central importance of diminishing returns to factors of production that can be accumulated, or the role of biased technological progress, is easier to understand in a fully aggregative context. But the way these principles work themselves out in practice may need to be studied in an explicitly multisectoral model. The same is true of basic demand-side influences, like different income elasticities of demand. Long-run variations in the composition of aggregate output are visible to the naked eye. They need to be understood.

This would hardly need saying to anyone who thinks of macroeconomics as a pragmatic discipline, not an ideological opportunity. In precisely that spirit I want to suggest one more currently salient issue about which an empirically validated structural dynamics would almost surely have useful things to say.

I have in mind a rather less clear-cut example. There are hints in the past decade or so that the wage share of national income may have fallen relative to the property share. Since the measured ‘wage share’ includes a return to human capital, and the human-capital component itself may have been increasing, the share of ‘raw’ labour may have decreased more sharply. In the US at least, however, the measurement of income shares has been strongly affected, and may have been distorted, by the growth of the financial services industry, which may be transitory rather than ‘structural’. Suppose, however, that there does indeed turn out to be something of a long-run character to explain.

Then once again a multisector, multi-factor growth model would be the natural vehicle. Relative shares in national income are weighted averages of sectoral shares in value added. Long ago I wrote an article that showed, using fairly primitive methods, that much of the apparent time-series stability of relative income shares could be accounted for simply by this averaging process. A full explanation would have to account both for shifts in the intra-sectoral distribution of value added and for endogenous and exogenous shifts in the sectoral weights. Something might need to be said about both factor substitution elasticities within sectors and demand-side substitution elasticities among sectoral outputs. There may be other, more purely macroeconomic, forces at work. Incorporating them in a multisector growth model would force an explicit formulation of any such forces that is in any case desirable.

I remember a joke from my childhood. Scholars of many different nationalities are shown an elephant and asked to write a brief essay about it. The German writes An Outline of an Introduction to a Treatise on the Fundamental Nature of the Elephant, the Frenchman writes on L'Elephant et l'Amour, the Pole writes on The Elephant and the Polish Question, and so on. The point I am trying to get across is that structural dynamics is a natural extension of ordinary economics and worth continued attention, not some exotic addendum to be attached to near-irrelevant special interests.

In his contribution to the present volume, Professor Solow has chosen to reproduce his own discussion at the Lincei Academy on my book (Pasinetti, 2007) with an ‘Afterword’, bringing out the link between the legacy of the Cambridge Keynesians and structural dynamics. I propose to do the same here. This seems to me to be the best way to respond to the Editors’ invitation to contribute, i.e. to offer my own ‘Reply’ to the discussion with an ‘Afterword’ appended to it.

To begin with, let me say how grateful I am to Professor Solow for taking the sting out of Joan Robinson's original ‘devastating insult’, as he has taken it, and proudly defining himself a ‘bastard Keynesian’. Admirable! But not yet enough, I think, to take full advantage of his generous stand. He says: ‘I don't remember that [Pasinetti] ever says . . . exactly what the Keynesian revolution was about and how it is incomplete, what the remaining gap is’, after stating: ‘Keynes's great work of 1936, is a difficult book because it contains several different themes. It is not always clear that the different themes are compatible with one another, so the book is very hard to summarize or to give to beginning students to read. I think Luigi Pasinetti's work shares some of that quality; there is more than one thing going on and it's not always clear – at least not always clear to me – exactly how the several themes fit together’ (see above).

I take this as a great compliment, for which I am thankful. It may incidentally also give some substance to my repeated remarks on the ‘lack of communication’ in the Cambridge Group. Yet I cannot take it in the literal sense. I think it is a very nice way of saying that he does not agree with me.

There are plenty of references to the ‘Keynesian Revolution’ in the book, as one may gather even simply by using the long list appearing in the Index (with all its variations, see Pasinetti, 2007: 375–6); or the Table of Contents, and particularly Chapter VIII, where there appears a whole specific section (‘The ideal task of Keynesian economics’, in ibid: 269–273) which is framed with the purpose of specifying the substance of what, in my view, the Keynesian Revolution should aim at. And all this, even without mentioning other relevant references, such as the three recent conferences¹, which both Professor Solow and I attended, where I presented papers precisely on this issue. So much have I been concerned, as to begin my Preface by immediately stating: ‘The “Keynesian revolution” only succeeded half-way . . . In terms of economic policy its successes were immediate [and these are indeed the aspects to which Professor Solow refers, in stating his view of the Keynesian Revolution, but] in terms of economic theory [and this is the part that Professor Solow leaves aside], Keynes's original ideas failed to achieve wide acceptance. Economic science essentially continued to do “business as usual”, i.e. with a Walrasian engine at its core’ (ibid: xiii). Most probably, the crucial discriminating point is precisely here (with the ‘Walrasian engine at its core’). This is what Hicks did, explicitly, in his immensely popular IS-LM Keynesian model (Hicks, 1937). It is again the point from where Franco Modigliani started (by his own explicit statement) in his famous ‘Liquidity Preference’ article (Modigliani, 1944). It is basically the view of the Keynesian Revolution that is presented by Professor Solow here. Incidentally, this may also explain why, in a rough language, so alien to Solow's kindness (but not unusual in Joan Robinson), the replacing of Keynes's theory with that of Walras, and simultaneously accepting the Keynesian unemployment economic policies – justified, by the various authors, with a whole series and variety of attempts² – could not but be perceived (by the daughter of a Major General!) as generating ‘bastard’ offsprings. Solow's kindness now eliminates the unpleasant aspect. But the disagreement remains. For me, a genuine ‘Keynesian revolution’ cannot be confined to temporary measures aimed at saving a sinking ship in times of mass unemployment. Much else should be at stake.³

This issue leads me straight to the second point of disagreement. This refers to my presentation of the historical development of economic thought as following a succession of two relevant, contrasting, paradigms: one associated with the investigations of the process of exchange (relying on subjective individual preferences and the principle of optimum allocation of given resources) and the other associated with the investigations of the process of production (and relying on objective technical relations in production and on the principle of human learning).

I originally developed this contrast in a work (Pasinetti, 1986) in which I investigated the ways in which the subjective and the objective theories of value (as opposed to pre-theoretical approaches) have emerged in the history of economic thought, and how they can be placed at the basis of the two mentioned (opposed) paradigms, in a way similar (though not entirely coincidental, due to our moving in the field of the social sciences) to that suggested by Thomas Kuhn (1970) in his well-known work on the way ‘scientific revolutions’ have taken place in the history of science.

I feel Professor Solow may have underestimated the importance of this distinction. I should like to stress that I have repeatedly and at length given my reasons for making it. Some authors, however, make, in my view, excessively accommodating concessions to the objections made to the above-mentioned distinction, thus risking to lose the enormous analytical interpretations and possibilities that such a logical framework makes possible.

I find it significant that Professor Solow should refer precisely to the formalization given by Jacob Mosak – namely the one which I myself originally used in my own work – to acknowledge the logic and clearness of a model of pure exchange. But he attributes it to being a device for didactical purposes. I think my counter-arguments on this point are much stronger. I have defined explicitly at the very beginning (Pasinetti, 2007: 19–20) the criterion that one should adopt to single out the features characterizing what can be called the basic model of each paradigm. As I specify, they should be those features, and only those features, that the basic model ‘cannot do without’. It is logic, not didactical expedience, that must be considered in order to define what is the basis characterizing the structure of each paradigm. And on this basis, the gap between the two paradigms appears astonishingly huge. I found it interesting to recall on this point how strongly Piero Sraffa felt, when commenting on the gap that separates the approach of the Marginalists from that of the Classics (the two major representatives of the two above-described, contrasting, paradigms). He used the words ‘abysmal gulf’ (ibid: 195).

Yet I did not confine myself to relying on arguments of logic only. I thought it necessary to look deeply into the facts and thus I thought it appropriate to insert factual arguments in the form of a rather extended excursus on the ‘historical background of economic analysis’ (ibid: 250–255), in which the clear emerging of the two paradigms, on the historical scene, is presented with a fully thought-out explanation and rationale.

Professor Baumol should have no fear on this point. There is no ‘prohibition’ on my part on the pursuit of other approaches to investigate particular aspects, provided they are inserted in the appropriate way. The production paradigm, unlike the pure preference paradigm, is developed with a plurality of degrees of freedom (see further hints below). There is no reason to reject one paradigm, as such, as ‘wrong’, with respect to the other. Both are given a justification, but with reference to different and appropriate historical phases in the evolution of economic analysis. And in the historical phase that we are living in at present, the pure preference paradigm has become terribly insufficient.⁴

One clear reason why it has become insufficient may indirectly be seen from the standard defence of mainstream economics (rational expectations included) displayed by Professor Bertola.⁵ This is surprising to me, at a time when most economists are having serious doubts about ‘modern economics’.⁶ It is even more surprising at a meeting like the present one, in which the defence of neoclassical theory has very wisely and explicitly been avoided also by Professor Solow (see the beginning of his contribution).

It is possible, to judge from what Giuseppe Bertola writes, that he may not have looked at the third part of my book with the concentration that it would have needed. He starts from the sub-title of the book and a bit too quickly tries to deal with it by an intelligent reference. Of course – he says – the Keynesian model is incomplete. It must be incomplete. ‘In 1931, Gödel proved that every logical system has to be intrinsically self-referential and incomplete: not only must something be assumed but much else must remain logically undecidable.’⁷ By all means, Gödel's proof was a remarkable step forward, in logic and philosophy. It refers to all logical systems as such, not – as in my case – to the way in which alternative logical systems may develop in history. However, I must say, I find Professor Bertola's citation a splendid one. I am only surprised that he does not seem to realize that it applies perfectly well to the general equilibrium model, which he is so innocently defending. The reason simply is that such a model is – or it is aimed to be – a closed model. And, as all closed (or would-be closed) axiomatic models, it is indeed subject to Kurt Gödel's famous undecidability proof. Had Professor Bertola paid a little more accurate attention to my Book Three – which is the most engaging part of my work – he would surely have realized how concerned I have been with the thorny problem of the complexity of the industrial world, for which I am presenting the newly proposed paradigm (Pasinetti, 2007: 274–79, 307, 323–30). He would surely have come across what I have called the separation theorem. By it, I make the proposal to separate ‘those investigations that concern the foundational bases of economic relations . . . from those investigations that must be carried out at the level of the actual economic institutions’ (ibid: 275). Only some readers have in fact paid explicit attention to this (I think) important aspect of my model. Baumol tries a rather eclectic attempt – We need them all – he says of all attempts. In principle, of course, I agree. I am not against any specific approach, provided that it is used appropriately. Even less am I advocating any prohibition! I have tried my best to sketch out a logical framework for ordering the investigations of complex phenomena (ibid: 267–69). Perhaps my reply to what I call the third objection, to my separation theorem, as is anticipated and specified in my book (ibid: 327–28), seems to go in Baumol's direction. The logical framework of what I call ‘the natural system’ – very much unlike that of the general equilibrium model – is not closed. Many things have to be decided from outside it – not because of Gödel's theorem, which does not concern this aspect, but owing to its own logical characteristics.

Let me finally come to a few brief comments on some specific questions. I was first asked to what extent, in my mind, do ‘natural’ and normative coincide. A complete answer would require going into long details, and my method of referring to specific pages of the book would not be so effective in this case. The shortest answer I can think of is to say that the two concepts overlap to a great extent, but do not coincide. It is easy to think of relations that are ‘normative’, but do not belong to the ‘natural system’, and equally well of relations that belong to the ‘natural system’ but are not necessarily (or exclusively) ‘normative’.

The second question is more intriguing. Professor Baumol was asked if he would be prepared to utilize the phrase ‘magnificent dynamics’, that he coined a long time ago, in connection with the paradigm which I proposed. Professor Baumol has avoided, perhaps wisely, picking up the question. But – interestingly enough – it is again Professor Solow who has become involved with it, even if in an indirect way. The term ‘magnificent dynamics’, as is well known, was coined by Baumol (1951) with reference to the grand dynamics of the Classics, of Marx, of Schumpeter and of Harrod (indeed the first economist to extend Keynes's theory to the long run). It is at least legitimate for us to begin by asking: does Solow's neoclassical growth model belong to Baumol's category? I think the answer is no, as Solow himself explains: ‘While it is true that the growth paths of capitalist economies are never paths of continuous full employment, full utilization equilibrium, it is true as a matter of historical fact that it would be reasonable to describe those paths as fluctuations around a trend, with most of the time the fluctuations contained within fairly narrow bounds – a . . . corridor around that path.’ This is a tricky proposition. It seems to me an attempt to get away from the criticism I made (Pasinetti, 2007: 231n), which points out that Solow's growth model (1956) is logically consistent only in a hypothetical world of proportional dynamics, i.e. in a world with a single commodity (or at most a composite commodity with absolutely constant composition through time), which – I point out – can never be the case in economic systems with technical progress. This may also be the point at which Professor Solow has come to realize the relevance of my work on structural dynamics, as he nicely states. I should be so glad to find him consistent one step further, by accepting that his proposition that ‘the Keynesian revolution will never be complete because behaviour and institutions in our society change. They evolve’ contains a contradiction. Revolution and evolution are antithetic terms. The real way to avoid these contradictions is to accept my separation theorem: a break (a revolution) with respect to neoclassical theory, and a set of investigations of the appropriate (evolving) institutions. Indeed, in agreement with Professor Solow, though in a different direction, I sincerely do ‘hope small armies of young economists will engage in trying to bring the Keynesian revolution’ to its accomplishment.

Nevertheless, at the end of this discussion, it is encouraging for me to realize that the two different directions just hinted at (Solow's and mine) may not be so entirely divergent as might at first appear, after all. It is comforting for me to discover that there is an overlapping area in our evaluations – the one concerning structural economic dynamics – on which we seem to entertain similar convictions as to its relevance, and therefore as to its being a really important field of research for younger generations of economists to concentrate on. I am sure Solow's position on this is so pre-eminent and inspiring for the armies of young economists as to make me more than happy to be able to join him.

I am glad Professor Solow has decided to participate in the present Collection of Essays by deciding to reproduce here his contribution to the recent Lincei Academy discussion on my latest book, and most of all by making a crucial addition to it – namely an afterword devoted to structural economic dynamics. This makes it only too logical (as explained above, p. 276) that I should myself follow his example by reproducing my ‘Reply to criticisms’ in the same discussion, and similarly by adding the present afterword on the same subject.¹

The question at stake is not a minor one. It is in fact a fundamental question, as it goes down to the very significance of the type of analysis that can be carried out with a one-sector-model approach associated with Solow's neoclassical growth model versus the significance of the structural economic dynamics approach associated with my own work.

The crux of the matter is revealed by the proposition with which Solow opens the substantial part of his afterword. Writes Solow: ‘I cannot imagine how anyone can be “against” the goal of structural dynamics. Multi-sector growth models and one-sector growth models are complements, not rivals. The main reason for pursuing one-sector and two-sector analysis is transparency. The role of certain fundamental principles . . . is easier to understand in a fully aggregative context. But the way these principles work themselves out in practice may need to be studied in an explicitly multi-sectoral model’ (see p. 274 this volume).

These sentences shed light precisely on the major point of contrast between Solow's one-sector model approach and the approach to structural economic dynamics which I have been pursuing in my works. There are, of course, many differences between the two approaches, because of the different theoretical paradigms from which they stem. But there is one major contrast on which I shall concentrate here – it concerns the specific topic of structural dynamics.

I want to argue that the structural dynamic approach to growth is not merely complementary to the one-sector approach. It goes much beyond complementarity, to such an extent that, when we consider the passage of time, the one-sector approach becomes incompatible with structural analysis.

Essentially, the two approaches embody two different visions of the industrial world. The vision behind structural dynamics originates from the consideration of a permanently evolving economic system. The vision behind the aggregate model of traditional growth theory embodies a static, or at most a stationary, view of the economic system, and the reason is that it is inherently incapable of absorbing any change in time of the structure.

At a first glance, a multi-sector model may look like a simple, more detailed, specification of the one-sector model. And hence the analysis of the structural dynamic type might at first appear as complementary to the traditional aggregate growth model. But in fact this apparent impression is false.

Any multi-sector model is characterized by having, at any given point in time, a particular structure. The size of the aggregate output of the economic system will, of course, turn out to be a weighted average of the outputs of the various sectors, the weights being given by the sectoral proportions, with respect to aggregate output, of the quantities of each good demanded as a consumption good and/or as an intermediate commodity for the production of (all) goods. This structure of the whole system derives from technology on the one side and from demand on the other side. But both technology and demand together will shape a structure which is specific to a specific point of time. When we introduce time, the whole structure will in general change; and as a consequence the whole complementarity between the two approaches will break down. Notice that it will break down on the side of the aggregate, not on the side of the structural, view.

We could, of course, imagine an economic system that grows in absolute terms, while maintaining relative quantities absolutely constant. In my 1981 book, I have shown that there are three highly hypothetical cases in which this can happen. I have called them the three cases of proportional dynamics. They are:

1. 1. The case of population growth at constant returns to scale.

2. 2. The case of uniform technical progress in all sectors and equally uniform expansion of demand for all goods and services.

3. 3. The case in which, in each single (vertically integrated) sector, the rate of growth of productivity is exactly equal to the rate of growth of per-capita demand, so that the two rates of growth cancel each other out inside each single sector.

These three cases are precisely those cases that must be taken as implicitly assumed by traditional economic theory when talking of growth, if we want to give logical consistency to such a theory. But such a type of growth would simply be an expansion at constant proportions (a kind of blowing up) of the (supposedly optimum) position assumed to have been achieved by the economic system, at the initial point of time. It would really be hard to attribute any significance to such a type of (proportional) dynamics. In fact, in such cases, it would not even make much sense to set up a multi-sector growth model, since we could simply compare total aggregate quantities, perfectly knowing that the whole structure within them remains constant.

If we are going to pay any attention to the criterion, which Professor Solow has always advocated, of empirical relevance, we should obviously rule out these three cases as failing to give any relevant representation of the real world. But if we rule out these cases, we are left with an economic system whose structure evolves through time, with productivity and demand growing at different rates from sector to sector and independently of one another. In this context, it will indeed make sense to talk of dynamics – of a true dynamics – in which not only the absolute but also, and most importantly, the relative magnitudes are continually evolving.

It is within a context of this type that the interaction of the unequal evolution of technology and of the inevitably complex evolution of the composition of consumption of goods and services gives rise to the structural dynamics of quantities, of prices and of sectoral employment. As a laborer, each single individual contributes to a very specialized part of the division of production tasks in each single sector, but as a consumer, he/she demands in principle the goods produced by all sectors of the whole economic system. Almost paradoxically, it can be shown that there exists a genuinely macroeconomic equilibrium condition that must be satisfied overall. Obviously, it will be the task of the institutions of the economic system to govern the interaction between individual efforts and consumption choices so as to drive the economic system to the fulfilment of such macroeconomic condition (which, as I have shown, turns out to be the equivalent of Keynes's effective demand/full employment condition).

I think I have been able to give the clearest version of such a macroeconomic condition by using a simple pure labor model. The results have been summarized in a few sentences which can be found in my 2007 book (p. 285), and which may be usefully synthesized as follows.

One must notice the essentially macroeconomic nature of the overall equilibrium condition, not so much because of the sum of the m sectors but more fundamentally because all sectors are linked – even when there are no inter-industry relations – by the effect of the overall demand. Each worker may contribute to a very tiny fraction of the production of a particular good (a sectoral contribution), but at the same time, with his/her family, will contribute to the demand for virtually all the goods and services produced in the economic system. Through this channel, owing to the all-embracing effect of overall demand, the set of all production processes forms a true economic ‘system’. Such a condition is indeed a single relation, but it concerns the whole economic system. In this way it becomes a single truly macroeconomic relation, independent of the number and structure of the sectors and irrespective of all the movements of both technology and of consumers’ demand.

I trust the reader will see, in this context, that the only approach capable of resuming a genuine dynamic analysis in economic growth models is not simply that of introducing time but of introducing time in a relevant way. This implies introducing a continually evolving structure of the economic magnitudes, as time goes on.

This is the gist and purpose of structural economic dynamics.