Lecture to the memory of Alfred Nobel, April 27, 1973

The Mainspring of Economic Growth

In my Theory of Wages, first published in 1932, there is a chapter (VI) entitled “Distribution and Economic Progress”. It was the first to be written of the theoretical chapters in that book; so it is in a sense the first of my contributions to economic theory. I do not think much of it now; I think that I have learned a good deal since I wrote it. It has nevertheless had a considerable progeny. Work that is based upon it, or on other constructions of the same character, continues to appear; so it is far from being dead. Yet I myself have moved away. It may be useful to take this opportunity of explaining how this has happened. It is not inappropriate to do so, since in doing so I shall describe what seems to me to be an important part of the work I have done, in all the time from 1932 to the present.¹

It was characteristic of that approach, from which I began, that it treated the Social Product as being made by two Factors of Production, Labour and Capital; the services of Labour and the services of Capital contributing to the Product in much the same way. An increase in the amount of either Factor that was applied would increase the Product, other things being equal. With given amounts of Factors applied, there would be just so much Product; so (again other things being equal) there would be a Production Function (as it later came to be called) representing Quantity of Product as a Function of Quantities of Factors applied. The return, per unit, to each Factor was equal to its Marginal Product, which diminished as the amount applied of that Factor increased, the amount of the other remaining constant. It followed at once that an increase in the quantity applied of one Factor (that of the other remaining unchanged) would increase the absolute share of the Product going to the other Factor; but since the absolute share of the increasing Factor might be either increased or diminished (according as its Marginal Productivity curve was elastic or inelastic) the distribution of the Product between the Factors (relative shares) might be shifted either way. Which way it went would depend upon the “shape” of the Production Function, a “shape” which could be represented, as I showed, by what I called the “elasticity of substitution”.

It was not supposed that the Production Function would remain unchanged over time; it would be shifted by the discovery of new techniques of producing – that is to say, by invention. Inventions, so Wicksell appeared to have shown (and I followed him), would not be adopted unless they raised the Social Product; but the shifts in the Production Function, due to invention, might be “neutral”, as far as distribution between the Factors was concerned, or might be biased either way. It seemed to me that rises in wages (rises, that is, in the share of the Product going to Labour per unit of Labour) would encourage the adoption of inventions which economised in Labour and so were biased against Labour; but whether such “induced inventions” were to be regarded as shifts in the Production Function, or as substitutions within an unchanged Production Function, was left rather obscure.

The theory, which I have been outlining, has been decidedly influential; but almost every element in it has been a target for criticism. Some of the criticism (such as those directed against the Marginal Productivity theory as such) can, I still believe, be rebutted, or partially rebutted;² but there is one that remains which I now feel to be decisive.

In the Production Function, “Product”, “Labour” and “Capital” are quantities; but it is necessary, if they are to be quantified, that there should be some means of reducing their obvious heterogeneity to some kind of uniformity. For none of the three is the reduction a simple matter; it cannot be solved, even in the case of Labour, by counting heads or by counting manhours. The crucial problem, however, is that of capital.³ Capital, here, must mean physical capital goods; it is an aggregate of physical goods which we have to represent by a single quantity. As is now well known (but was not so well known in 1932) there are just two cases in which this can be done without error – without any error, that is, for it is not denied that if either case is approached, without being actually reached, the error may be tolerable. One is the obvious case in which all components change proportionately; the other – which I myself may claim to have clarified in 1939 – is that in which the price-ratios between the goods, or their marginal rates of substitution, remain constant.⁴ In the former case the complex is representable by a number of physical “bundles”; in the latter there is aggregation in value terms.

It is clearly impossible, in the case of the capital stock, to claim that the first of these conditions, in practical application, can be even approximately satisfied. For it is normal experience, in a progressive economy, that its capital, at the end of a period, contains different kinds of goods from those contained at the beginning. New items are introduced, and old items discarded. Only in a theoretical construct – a steady state – can proportions remain unchanged over time; and we can hardly make much use of that property, even for the comparison of steady states, since proportions in one steady state will usually be different from those in another. There is little hope for a way out in that direction.

The other, at first sight, looks more appealing. Here however there is a more subtle objection, associated in particular with the work of Joan Robinson.⁵

If capital increases relatively to labour, other things being equal (so the Production Function theory appears to tell us) the marginal product of capital must fall, so the rate of return on capital must fall. But a fall in the rate of return on capital carries with it a fall in the rate of (real) interest, as a result of which the capitalised values of different goods (goods of different durability, for instance) must change disproportionately. So the marginal rates of substitution between them cannot be kept constant. The constant-price condition cannot be maintained; it involves a contradiction.

This does not mean that it is wrong for statisticians to value capital goods at constant prices – their prices, or costs of production, at some base date. Any practical measure of National, or Social Capital must I think be of this character. But a technological relation between Capital and Product, with Capital thus arbitrarily valued, carries no conviction; there is no reason why it should exist.⁶

All I have said so far is by way of preface. I have, on the whole, left it to others (since my early days) to live in the world of production functions and elasticities of substitution, between Factors globally defined. What I mainly want to talk about is a side of my work which has gradually developed over the years, and which, I now feel, is more promising.

It also goes back to that same Theory of Wages. I have so far been discussing Chapter VI; but there are other chapters (IX-X) where will be found the beginnings of quite a different theory. These are curious chapters; their reception, when the book appeared, was much less favourable than that then accorded to Chapter VI. This was partly because the tradition in which I was working in IX-X – the tradition of Böhm-Bawerk and Wicksell – was much less familiar to English readers than that of Pigou, on whom I was drawing in VI; but mostly because of a head-on collision between what I was saying and the “New Economics” which even then, three years before the General Theory, was already beginning to be the Economics of Keynes.

When I wrote the Theory of Wages, I was completely innocent of these ideas; I had scarcely a notion of what was going on at Cambridge, or for that matter in Sweden. But hardly had my book left my hands when I began to move in that direction myself. I stumbled upon something which, if not quite the same as Keynes’s Liquidity Preference, has a close relation to it. And even before the General Theory appeared in 1936, I had begun to draw some of the consequences.⁷

There is much of my work which follows from that; I have no time to describe it here. I must keep firmly to the story of those chapters in the Wages book, and what follows from them. The first result of the new point of view, when I reached it in 1933-35, was to make me deeply ashamed of what in those chapters I had written. I realised (too late) how inappropriate it was. It had nothing to do with the state of the world at the time when I was writing. I had diagnosed a disease, but it was not the right disease. The unemployment of 1932 was of quite a different character from what I had supposed.

It is nevertheless not useless to analyse a disease, even if it is not the disease which at the moment is important. The time may come when one’s work is more to the point. In my case, I think, it has come.

The principal ground on which my chapters were attacked, in the thirties, concerned my initial assumption – that Trade Unions, or Government wage-fixers, can raise real wages. This, by Keynes and his followers, was in those days most resolutely denied. Trade Unions, they said, are concerned with money wages, not real wages. It is true that a rise in the money wages of a particular group will raise their real wages relatively to those of others; but a general rise in money wages, in a closed system, will simply result in a rise in prices in the same proportion, thus leaving real wages where they were. This of course implies that there is an elastic money supply. If the money supply is not increased proportionately, the rate of interest will rise; as a result of the rise in interest there will be a fall in the demand for labour. The cause of the unemployment is then identified as the inelasticity of the money supply.

It is fairly obvious, in these days, that this Keynesian argument is not so strong as it at first appeared. Directed, as it was of course at first intended to be directed, against the use of wage-cuts as a means of stimulating employment in depression, it retains its force. But it is much less strong on the other tack.

Though Trade Unions operate on money wages, it is surely in real wages that they are really interested. If a rise in money wages just leads to a rise in prices, they feel themselves cheated; so they return to demand another round of rises in money wages. Thus we get the cost-inflation, with which (during the last twenty and especially ten years) we have become so familiar. It could not occur without an elastic money supply; so why not put constraints on the money supply, and so check, or at least impede, the inflation? There are bound to be monetarists who will argue that way, and governments, in desperation, are bound to give some attention to them. Is the resulting unemployment then due to the monetary constraint, or to the wage-push which led to the monetary constraint being imposed? One can look at the matter either way, but it can well be argued that the latter way is the more fundamental.

So my 1932 analysis has come, at last, to some sort of contemporary relevance; but there is another kind of relevance, of which I had no suspicion when I wrote, but which has been there all the time. This is not a matter of analysing a disease; it is concerned with the normal growth, the healthy growth, of an economy. In healthy growth real wages should be rising. What are the consequences of that rise in real wages? My 1932 analysis was concerned with rises in wages off the normal path; but the rises that are on the normal path should have similar effects, though they will not include the causation of unemployment. Rather similar methods should be usable for their analysis; it should deepen our understanding of the growth process in general. This is the aspect in which I have lately been mainly interested. I will try to sketch some of the results I seem to have been reaching. This will be the subject of the rest of this lecture.

I have talked all this time about those chapters IX-X of the Wages book, without specifying what they contain. Much of what they contain is detail, now irrelevant. There is just one thing that matters.

A rise in real wages, however caused, tends in itself to diminish the real rate of profit. This has two effects which work, in a sense, in opposite directions. One is to encourage the substitution of what are usually more capital-intensive methods; the other, because of the transfer of income from profits to wages, is to diminish saving. Far more is now known about both of these effects than I knew in 1932; but the distinction still holds. I will try to re-state it in a more modern form.

The first thing on which to insist is that it is quite unnecessary, because we use terms like “capital-intensive” and “rate of profit”, to trouble ourselves about the valuation of the capital stock as a whole (as we appeared to have to do on the production function method). What matters is not the average rate of profit on the whole capital stock (which cannot be determined without such valuation); what matters is the rate of profit on new investment. When the new investment is undertaken, that profit is no more than an expected profit, and what is realised may not be the same as what is expected. It seems reasonable, however, if we are concerned with healthy growth, to suppose that there is some broad concordance between what is expected and what is realised. Most ventures come out more or less right. No more than that is required.

It cannot be profitable (in this sense) to make machines unless the use of the machines is also profitable; so, to assess the profitability of investment, we should look right forward to the production of final product. In any production plan, so considered, labour is input and final product is output; so a rise in wages, in terms of final product, must diminish the rate of profit on the plan, in terms of final product. To this rule there is I believe no exception. It holds for any plan that could be viable at the rate of wages in question.⁸ So for any plan (with inputs and outputs expressed in quantity terms) there is a particular relation between (real) wage-rate and (real) profit-rate, which can be drawn out as a downward-sloping curve – what I now like to call the “efficiency curve” of the plan.

Next (though only provisionally) let us make the conventional assumption that “technology” is given; that there are just so many production plans, in the above sense, from which choice can be made. Each such plan will have an efficiency curve. Make the “capitalist” assumption (I am not here concerned with the question of its justification) that the plan which is actually chosen for new investment is that which gives the highest return at the current rate of wages. It could be that the choice was unaffected by the level of wages; but it makes more sense to suppose that as wages change, different plans (or techniques) will come to be the most profitable. There will then be substitution along a “spectrum of techniques” as wages rise.

It is not the case (as used to be supposed) that there is any single physical index by which we can distinguish those techniques which lie “further down” the spectrum from those which lie “higher up”. There is no such index which can be employed without exception. I could already show (in 1939) that the “Period of Production” that was used for this purpose by Böhm-Bawerk and Hayek will not in general serve.9 But what is in substance the same argument can be used against any physical index, such as capital-labour ratio (when capital, by some device, is physically defined).¹⁰ Yet we should not allow these refinements to obscure the fact that techniques which lie further down the spectrum (so that they require for their profitable adoption a low rate of profit, or interest) will usually be such as to involve higher preparatory costs, such as construction costs, as a means of economising in running costs of production. We do not usually go astray if we think of such techniques as being in that simple sense more capital-intensive.

I shall later return to this substitution effect; for the moment I turn to the other, which is more troublesome. In Theory of Wages (as was natural at the date when it was written) I took the traditional view that more saving meant more capital accumulation, and that capital accumulation was favourable to rising wages. But in Keynes’s system of thought, which was so soon to be sprung upon me, the effect of saving seemed to go the other way round.

The trouble was not (as might easily be supposed) that Keynes’s theory was monetary, while my “classical” theory was non-monetary. One can construct a “barter” system, in which money plays no essential part, but which can still behave in the manner that Keynes identified. (It is not, incidentally, such an unrealistic construction; the world, in 1970-1, produced quite a good imitation of a Keynesian slump in real terms.) It has taken some time for this to be clarified – since Keynes himself, by unfortunate definitions which made saying and investment always equal, obscured the significance of a part of what he was saying.

If we make the distinction (which already in 1936 was familiar in Sweden) between desired and realised saving and investment, the issue becomes much clearer. In the desired sense there can be an excess of saving over investment, even in a barter economy; it will take the form of an undesired accumulation, an accumulation of surplus stocks. If there is an excess of investment over saving, in the desired sense, stocks will fall below normal, below what is desired; or surplus orders will pile up, orders which cannot be satisfied without abnormal delay. Such an excess, either way, may be regarded as a sign of disequilibrium – a disequilibrium which is perfectly possible, even in a barter economy.¹¹

Saving-investment equilibrium, so defined, does not imply the Full Employment of Labour; for that also to be attained, further conditions are necessary. One of the conditions is that relative prices should be right. It is unnecessary, here, to discuss the vexed question whether it will always be possible, in a barter system with sufficiently flexible prices, to maintain both full employment and saving-investment equilibrium automatically. (I am myself convinced that it is not necessarily possible, but that is by the way.) What is important, for my present purpose, is that saving-investment equilibrium and full employment are different. One can suppose that there is saving-investment equilibrium, maintained continuously; and yet there can be unemployment, if the ratio of prices to wages is inappropriate. That is what I ought to have said in Theory of Wages. So interpreted, the Keynesian view and the “classical” view fit together.

It has taken a long time to clear this up. In the central part of my Contribution to the Theory of the Trade Cycle (1950) I used what I have later called a fixprice model.¹² I introduced an equilibrium path – a saving-investment equilibrium path – and a full employment path which lay above it. I was interested only in departures from equilibrium; so the only function attributed to the full employment path was to act as a Ceiling, which imposed a constraint upon the disequlibria which could occur. I did not ask why the equilibrium path should lie below the Ceiling. Indeed, I said much too little about each. I just drew them as straight lines – which is a simple way of saying nothing about them!

The natural way of finding out more about them is to consider the possibility of maintaining both saving-investment equilibrium and full employment. Suppose that both conditions have to be fulfilled; what will be the consequences? Real wages, it is clear by now, will have to be flexible; can they be kept flexing always upwards? If they can, it may be that the simultaneous satisfaction of both conditions can be maintained without friction (this does not mean that it must be attained); if they cannot, if there must be fluctuations in real wages when both conditions are satisfied, there will surely be greater difficulties. To learn more about the double-equilibrium path (as we may call it) seems thus to be the next thing required.

It has been widely appreciated that it is the next thing required. Many (though by no means all) of the “growth models” that have been developed on all hands during the last twenty years can be considered as answers, or attempted answers, to the question just put. Some of them, especially those labelled “neo-classical”, use the production function scheme I began by describing. I am myself untempted by that procedure, essentially for the reason given. It may nevertheless be agreed that the problem is a “classical” problem; since we are putting disequilibrium behind us, what we have learned from Keynes is for the moment irrelevant. It is to the classics that we must go for help. We shall find it, in my view, not in the “neo-classics” but in the British Classical Economists, especially in John Stuart Mill.

When Mill “abandoned the Wage Fund” he must have forgotten what he had said about it. (It is not surprising, in view of all the other things he had been doing, if by 1868 his recollection of his earlier work had become a little rusty.) In terms of the double equilibrium path, what is said in his Principles is substantially right. The wage-bill (the real wage-bill) is just the difference between final product and what is taken out of that product for other purposes. What is taken out will include not only “consumption out of profits” but also the consumption of public bodies (as Adam Smith, when he was on this track, had been well aware). So long as the increment in this Take-out does not exceed the increment in final product, the real wage-bill must increase when final production increases. It must do so, along the double equilibrium path.¹³

We can at last begin to see how the substitution effect and the saving effect fit together. It is essential to hold fast to the behaviour of final output. This is the chain of causation: from investment to final output, from final output to wages, from wages to the rate of profit on new investment, and thence back on investment itself. There is much to be said on each of these steps. I cannot go into detail, but must confine myself to giving a general impression.¹⁴

Let us start from the making of an invention, which we had better think of as a major invention, so that the technique of production which it makes possible is much more profitable than any used before.¹⁵ It needs to be embodied in new equipment in order that it should be used; so without new investment it cannot be applied. But even if the invention had not been made there would have been some new investment; so the immediate result of the invention is that the technique which is embodied in new investment is changed. The rest of the economy proceeds more or less as before, using old techniques; they are now obsolescent, but they cannot be changed overnight.

The new processes will not produce final product at once; there must be a delay before the new equipment comes into production. During that delay, all final product comes from old processes; so (in double equilibrium) the old processes must continue in full production if final product is not to fall. Except by a fall in final product, no additional resources can be transferred to new investment; so it is just the resources which would have been employed in new investment, if the invention had not occurred, which can be transferred to the making of the new “machines”.

It is by no means certain that final product will be increased even when the new equipment comes into production. For it may well be that the increased profitability of the new machines is simply a matter of reduction in running cost. They have no larger capacity than the machines they replace; it simply costs less to run them. There is then no rise in final output when the new machines come into production. What does happen is that resources are released; but if double equilibrium is to be maintained, they must still be employed. They may be employed in squeezing additional output out of old processes; or they may be employed in making new machines. In the former case, there will at that stage be a rise in final output; and even in the latter case, though there is again no increase in output while the extra machines are being made, there will in the end be an increase in final output. So it is true (as Wicksell supposed) that a profitable invention will always lead to an increase in final output; but it is perfectly possible that the increase may be long deferred.

Unless the rise in final output is absorbed by an increase in Take-out, or has to be spread over too large an increase in the supply of labour, rising final output (when it comes) will mean a rising rate of real wages. (This is where substitution will come in.) But suppose for the moment that there is no substitution and no further invention. Investment continues on the same pattern as was established after the first invention occurred. Gradually, as old machines are replaced, the part of the capital stock which has become obsolescent will diminish; more and more will be of a “modern” type. During all that time final product will be expanding, and wages rising. As wages rise, the rate of profit will decline, from the exceptional level reached just after the original invention, towards something more “normal”. It will decline, though not before the modernisation has been completed, to the level which is appropriate to a steady state under the new technique; for the level of wages which is established in that steady state is the highest that can be achieved (except by diminishing Take-out) so long as there is no substitution and no new invention.

We need not rely, to establish this conclusion, on the “classical” view that a declining rate of profit will diminish the incentive to save.¹⁶ Whatever be the nature of saving propensities, an approximation to a steady state is likely to occur, if there is no further technical change. It will be a different steady state, with a different distribution of income, according as saving propensities take one form or another. But it will always be a steady state; and in that state wages will always be higher and profit lower than they were on the way to it.

Now we can bring in substitution. If there is substitution along a spectrum of techniques – new techniques which would previously not have been profitable becoming profitable because of the rise in wages – the fall in profits will be slowed up. The effect of the substitution (in most cases at least) will be in the direction of adopting more capital-intensive techniques. These will probably, at their adoption, slow up the rise in final product; and that probably means that they will slow up the rise in the rate of real wages. (Since they are directed towards economising in labour, it is not surprising that they should slow up the rise in wages.) But – and this is vital – the result of the substitution will be to set the economy “aiming” at a steady state with a higher final product per unit of labour, and therefore (with any reasonable behaviour of take-out) a higher level of wages. There are several ways of establishing this essential proposition;¹⁷ the simplest, perhaps, is just to observe that in the steady state, when the system is fully adjusted to the new technique, every worker will have more “capital” to help him when the method of production is more capital-intensive.

What I have just been giving is no more than an exercise; it does no more than distinguish one causal sequence which in actual experience will be crossed and mixed up with many others. It does nevertheless appear that this sequence may be rather fundamental. The mainspring of economic progress, it suggests, is invention; invention that works through the rate of profit. Each invention gives an Impulse, as we may call it; but the Impulse of any single invention is not inexhaustible. The exhaustion is marked by falling profit; but the cause of the exhaustion (on the Full Employment, or double equilibrium path) is scarcity of labour.

In saying this we are keeping, in substance, quite close to Mill. In Mill the “declining rate of profit” is due to scarcity of land; but there is no reason in principle why the operative scarcity should not be any natural scarcity. Mill’s view that the operative scarcity was land scarcity can only be regarded as empirical; it looked like being right at the time when he was writing, but over the whole time since then it does not look like being right. The ultimate scarcity must be that of labour or of land (or both); formally, it must be scarcity of some non-augmentable factor of production. There are of course many other scarcities which will arise in the working – out of the Impulse; but scarcities that can be overcome by investment will not reduce the rate of profit on new investment in general. They will shift the point in the productive process where the investment is to be made, but that is all. It is only the irremovable scarcities which will ultimately compress the rate of profit.

Once we recognise that substitution, on the spectrum of techniques, is just one way of overcoming the scarcities that arise out of the Impulse, many things fit into place. If there is no technical change, following on the original invention, other than that which is directly implied by the invention, the Impulse which it gives will soon be exhausted. Scarce factors will then get the full gain which accrues to them from the original invention, but no more. But if there is substitution, directed towards economising in those same scarce factors, the ultimate gain to them will be greater, and may well be far greater. There may still be a question of how the gain is distributed between them. Taken together, however, they must gain in the end from the deferment of exhaustion.

It is of little importance, from this point of view, whether the substitutions are supposed to take place along an unchanged “technology frontier”, or whether they themselves partake of the nature of invention. The “technology frontier”, useful as it has been in the formation of the theory, and still (perhaps) indispensable in the first stages of presentation, is in the end a piece of scaffolding, that we can take down. The puzzles about “induced invention” then give no trouble. We have just to define them as technical changes, the possibility of which is newly discovered in the working – out of the Impulse, and which are such that it would not be profitable to make them until the scarcities by which they are “induced” have developed. They thus appear as secondary inventions, “children” of the original invention, its “economic children”; for we may surely allow it to have other “children” – technical children, “learning by doing” in a most extended sense – as well. The economies of scale, on which one school of economists lays such great stress, may well be introduced in much the same place. The great inventions will give great and long-lasting Impulses, because they have many “children”, of all these kinds.

It will clearly be difficult, in relation to contemporary experience, to draw a firm line between the primary invention and its “children”, when the latter are so broadly defined. Where the distinction is drawn is bound to be a matter of judgement, or of taste. In relation to earlier ages, where claimants to primary status are less thick upon the ground, the distinction may be easier.¹⁸ One can certainly detect, in the nineteenth century, one major invention that gives a recognisable, and separable, Impulse – the railway. The Railway Age was an Impulse, the working – out of which is clearly visible, since there seem to have been at least a couple of decades in which no Impulse of comparable magnitude followed. Many economists (including sometimes Keynes) thought the Depression of the nineteen-thirties to be a Pause of similar character. That could be, but it does not have to be, since the Disequilibrium of the thirties can well be explained in other ways. It is impossible to tell its story without laying great stress upon the monetary aspect, which I have been disregarding; it may well be that it is right to tell the whole of its story in those terms. Yet there may be something more. It would be a great help if we knew, better than we do, if there was something more; for it would help us to understand the innovative process, as it works in this century, and so to know, better than we do, how far we can count upon steadiness in the flow of innovation. The study of past Impulses, with the aid of a better classification, might well throw much light upon this vitally important matter.

I return, in conclusion, to the Theory of Wages problem – the consequence of maintaining a level of real wages which is higher than that which is appropriate for double equilibrium. After what has been said, we need not conceive of this problem in a static manner. The equilibrium wage-level may be allowed to be rising, but the actual wage-level is kept, all the time, somewhat above it. That this is a realistic problem cannot nowadays be denied; for the means that are available for the enforcement of such a wage-level are far more extensive than they were in the past. Granted this kind of a wage policy, what happens?

There seem to be two main cases. It is possible, in the first place, that the higher wage might be matched by a lower take-out. A lower take-out would raise the (real) wage-level that was consistent with double equilibrium; so the lower take-out should permit of the higher wage being attained without unemployment. So far, so good; it should however be noticed that if the course of the wage-level is established arbitrarily, fluctuations in take-out will probably be necessary in order to keep that arbitrary wage-level consistent with double equilibrium. The matter cannot be settled in this way once and for all.

Secondly, suppose that the higher wage-level is not matched by lower takeout. I do not think it can be doubted that this also is a practical problem. For when we remember how much of the consumable product of modern economies is “taken out” for social purposes, the demand for which comes from much the same source as the demand for higher wages, we must surely recognise that the alleviation which can come from lower take-out is likely to be limited. Say then that all that can be done in this direction has been done. Say also that saving-investment equilibrium is to be preserved. (It will not be easy to preserve it, but most of what is to be said under that head is well known; I need not enlarge upon it here.) What, under these conditions, will be the course of the economy?

The higher wage, as we have seen, will affect the techniques that are chosen for new investment; we may take it that they will, on the whole, be more capital-intensive than they would otherwise have been, at corresponding dates. Such techniques will in the end raise final output, per unit of labour employed, more than it would have been raised by less capital-intensive investment. But, all along, the volume of investment (in saving-investment equilibrium) will be lower than it would have been otherwise. Thus, although the economy is “aiming” at a steady state in which final consumable output, per unit of labour employed, is higher, employment along the path to that state (and – in principle – even in that final state) will be lower than it would have been. This, I now believe, is what I was trying to say in Theory of Wages; as will be seen, it is subject to many qualifications of which, when I wrote, I had no idea. But in substance the main point stands.

That it is possible for a “developing” country, by choice of techniques that are too capital-intensive, to expand employment in its “modern” sector less rapidly than it might have done, is nowadays familiar. What I am saying is little more than an application of that same principle.

1. As seems suitable in this place, I shall confine this account to the evolution of my own ideas, without much attention to what I have learned from others. How much I have learned from others – especially, perhaps, from Roy Harrod, Joan Robinson and Nicholas Kaldor – will nevertheless, I hope, be apparent.

2. As I have explained in the Commentary attached to the second (1962) edition of Theory of Wages, especially pp. 333-41.

3. Crucial in the sense that it has been the major theme of controversy among economists. I accept that the aggregation problems, on the side of Product, are hardly less pressing.

4. Value and Capital (1939), p. 33 and passim.

5. The Production Function and the Theory of Capital (Review of Economic Studies 1954); and later writings.

6. I have stated my present views on the Production Function at greater length in Capital and Growth (1965) pp. 293-305; and in Capital and Time (1973) pp. 177- 84.

7. A Suggestion for Simplifying the Theory of Money (Economica 1935, reprinted in my Critical Essays in Monetary Theory, 1967). I have told the story of my “conversion” in Recollections and Documents (Economica, February 1973).

8. Capital and Time (1973) ch. 2.

9. Value and Capital ch. 17.

10. This is the principal point which the “re-switching” controversy is about. For further detail, see Capital and Time, ch. 4.

11. The disequilibrium which in a closed economy is revealed by physical stocks is in an open economy mainly revealed by foreign exchange – the balance of payments. For foreign exchange, in a modern national economy, is the easiest stock to run down, or pile up.

12. See especially Chapter 8 of the book. Also Capital and Growth (1965) chapters 7-l 1.

13. On Mill, see Capital and Time, pp. 58-62.

14. There is a much fuller discussion in Capital and Time, chs 9-10.

15. In the case of an open economy, the opening of a new market for an export, or potential export, will have a similar effect.

16. Nor even upon the sophisticated re-statement of the “classical” view that is due to Cassel (Nature and Necessity of Interest).

17. My own preferred way is that which is set out in Capital and Time ch. 10.

18. These ideas were not fully formed when I wrote A Theory of Economic History (1969), but I was working towards them.