According to this, Marshall's generalisation holds for the periods from 1880 to 1884 and from 1884 to 1886, but for no subsequent periods. It seems that we have been living all these years on a generalisation which held good, by exception, in the years 1880-86, which was the formative period in Marshall's thought in this matter, but has never once held good in the fifty years since he crystallised it! For Marshall's view mainly prevailed, and Foxwell's contrary opinion was discarded as the heresy of an inflationist. It is to be observed that Marshall offered his generalisation merely as an observed statistical fact, and, beyond explaining it as probably due to wages being stickier than prices, he did not attempt to support it by à priori reasoning. The fact that it has survived as a dogma confidently accepted by my generation must be explained, I think, by the more theoretical support which it has subsequently received.
To my statement that Marshall's generalisation has remained uncorrected until recently there is, however, an important exception. In his Industrial Fluctuations, published in 1927, Professor Pigou pointed out (p. 217) that 'the upper halves of trade cycles have, on the whole, been associated with higher rates of real wages than the lower halves,' and he printed in support of this a large scale chart for the period from 1850 to 1910. Subsequently, however, he seems to have reverted to the Marshallian tradition, and in his Theory of Unemployment, published in 1933, he writes (p. 296):
In general, the translation of inertia from real wage-rates to money wage-rates causes real rates to move in a manner not compensatory, but complementary, to movements in the real demand function. Real wage-rates not merely fail to fall when the real demand for labour is falling, but actually rise; and, in like manner, when the real demand for labour is expanding, real wage-rates fall.About that time M. Rueff had attracted much attention by the publication of statistics which purported to show that a rise in real wages tended to go with an increase in unemployment, Prof. Pigou points out that these statistics are vitiated by the fact that M. Rueff divided money wages by the wholesale index instead of by the cost-of-living index, and he does not agree with M. Rueff that the observed rise in real wages was the main cause of the increased unemployment with which it was associated. But he concludes, nevertheless (p. 300), on a balance of considerations, that 'there can be little doubt that in modern industrial communities this latter tendency (i.e. for shifts in real demand to be associated with shifts in the opposite sense in the rate of real wages for which work people stipulate) is predominant'.
Like Marshall, Prof. Pigou based his conclusion primarily on the stickiness of money wages relatively to prices. But my own readiness to accept the prevailing generalisation, at the time when I was writing my General Theory, was much influenced by an à priori argument, which had recently won wide acceptance, to be found in Mr R. F. Kahn's article on 'The Relation of Home Investment to Employment,' published in the Economic Journal for June 1931. The supposed empirical fact, that in the short period real wages tend to move in the opposite direction to the level of output, appeared, that is to say, to be in conformity with the more fundamental generalisations that industry is subject to increasing marginal cost in the short period, that for a closed system as a whole marginal cost in the short period is substantially the same thing as marginal wage cost, and that in competitive conditions prices are governed by marginal cost; all this being subject, of course, to various qualifications in particular cases, but remaining a reliable generalisation by and large.
I now recognise that the conclusion is too simple, and does not allow sufficiently for the complexity of the facts. But I still hold to the main structure of the argument, and believe that it needs to be amended rather than discarded. That I was an easy victim of the traditional conclusion because it fitted my theory is the opposite of the truth. For my own theory this conclusion was inconvenient, since it had a tendency to offset the influence of the main forces which I was discussing and made it necessary for me to introduce qualifications, which I need not have troubled with if I could have adopted the contrary generalisation favoured by Foxwell, Mr Dunlop and Mr Tarshis. In particular, the traditional conclusion played an important part, it will be remembered, in the discussions, some ten years ago, as to the effect of expansionist policies on employment, at a time when I had not developed my own argument in as complete a form as I did subsequently. I was already arguing at that time that the good effect of an expansionist investment policy on employment, the fact of which no one denied, was due to the stimulant which it gave to effective demand. Prof. Pigou, on the other hand, and many other economists explained the observed result by the reduction in real wages covertly effected by the rise in prices which ensued on the increase in effective demand. It was held that public investment policies (and also an improvement in the trade balance through tariffs) produced their effect by deceiving, so to speak, the working classes into accepting a lower real wage, effecting by this means the same favourable influence on employment which, according to these economists, would have resulted from a more direct attack on real wages (e.g. by reducing money wages whilst enforcing a credit policy calculated to leave prices unchanged). If the falling tendency of real wages in periods of rising demand is denied, this alternative explanation must, of course, fall to the ground. Since I shared at the time the prevailing belief as to the facts, I was not in a position to make this denial. If, however, it proves right to adopt the contrary generalisation, it would be possible to simplify considerably the more complicated version of my fundamental explanation which I have expounded in my General Theory. My practical conclusions would have, in that case, à fortiori force. If we can advance farther on the road towards full employment than I had previously supposed without seriously affecting real hourly wages or the rate of profits per unit of output, the warnings of the anti-expansionists need cause us less anxiety.
Nevertheless, we should, I submit, hesitate somewhat and carry our inquiries
further before we discard too much of our former conclusions which, subject
to the right qualifications, have à priori support and have
survived for many years the scrutiny of experience and common sense. I
offer, therefore, for further statistical investigation an analysis of the elements
of the problem with a view to discovering at what points the weaknesses
of the former argument emerge. There are five heads which deserve separate
First of all, are the statistics on which Mr Dunlop and Mr Tarshis are relying sufficiently accurate and sufficiently uniform in their indications to form the basis of a reliable induction?
For example, in so recent a compilation as the League of Nations World Economic Survey 1937-38, prepared by Mr J. E. Meade, the traditional conclusion receives support, not on à prioriecently available statistics. I quote the following from pp. 54-55:
During the great depression after 1929, the demand for goods and services diminished, and in consequence the price of commodities fell rapidly. In most countries, as can be seen from the graph on p. 52, hourly money wages were reduced as the demand for labour fell; but in every case there was a greater fall in prices, so that hourly real wages rose. . .[It is then explained that the same was not true of weekly wages.]. . .Since the recovery, the opposite movements may be observed. In most countries, increased demand for goods and services has caused commodity prices to rise more rapidly than hourly money wages, and the hourly real wage has fallen. . .In the United States and France, however, the rise in money wages was so rapid between 1936 and 1937 that the hourly real wage continued to rise. . .When real hourly wages are raised¾i.e. when the margin between commodity prices and the money-wage cost becomes less favourable¾employers are likely to diminish the amount of employment which they offer to labour. While there were, no doubt, other influences affecting the demand for labour, the importance of this factor is well illustrated by the graph on p. 53. In the case of all the countries represented for which information is available, the fall in commodity prices between 1929 and 1932 caused a rise in the hourly real wage, and this was accompanied by a diminution in employment. . .(it is shown that on the recovery there has been a greater variety of experience). . .This authoritative study having international scope indicates that the new generalisations must be accepted with reserve. In any case Mr Tarshis's scatter diagram printed below [in the Economic Journal, March 1939] (p. 150), whilst it shows a definite preponderance in the south-west and north-east compartments and a high coefficient of association, includes a considerable number of divergent cases, and the absolute range of most of the scatter is extremely small, with a marked clustering in the neighbourhood of the zero line for changes in real wages; and much the same is true of Mr Dunlop's results. The great majority of Mr Tarshis's observations relate to changes of less than 1.5 per cent. In the introduction to his Wages and Income in the United Kingdom since 1860, Prof. Bowley indicates that this is probably less than the margin of error for statistics of this kind. This general conclusion is reinforced by the fact that it is hourly wages which are relevant in the present context, for which accurate statistics are not available. Moreover, in the post-scriptum to his note, Mr Tarshis explains that whilst real wages tend to move in the same direction as money wages, they move in the opposite direction, though only slightly, to the level of output as measured by man-hours of employment; from which it appears that Mr Tarshis's final result is in conformity with my original assumption, which is, of course, concerned with hourly wages. It seems possible, therefore, taking account of Mr Meade's results, that I may not, after all, have been seriously wrong.
Furthermore, for reasons given below, it is important to separate the
observations according as the absolute level of employment is distinctly
good or only mediocre. It may be that we can analyse our results so as
to give two distinct generalisations according to the absolute level reached
by employment. If, at the present stage of the inquiry, we are to make
any single statistical generalisation, I should prefer one to the effect
that, for fluctuations within the range which has been usual in the periods
investigated which seldom approach conditions of full employment, short-period
changes in real wages are usually so small compared with the changes in
other factors that we shall not often go far wrong if we treat real wages
as substantially constant in the short period (a very helpful simplification
if it is justified). The conclusion, that changes in real wages are not
usually an important factor in short-period fluctuations until the point
of full employment is approaching, is one which has been already reached
by Dr Kalecki on the basis of his own investigations.
It may be that we have under-estimated the quantitative effect of a factor of which we have always been aware. Our argument assumed that, broadly speaking, labour is remunerated in terms of its own composite product, or at least that the price of wage-goods moves in the same way as the price of output as a whole. But no one has supposed that this was strictly the case or was better than an approximation; and it may be that the proportion of wage-goods, which are not the current product of the labour in question and the prices of which are not governed by the marginal cost of such product, is so great as to interfere with the reliability of our approximation. House-rent and goods imported on changing terms of trade are leading examples of this factor. If in the short period rents are constant and the terms of trade tend to improve when money wages rise and to deteriorate when money wages fall, our conclusion will be upset in practice in spite of the rest of our premisses holding good.
In the case of this country one has been in the habit of supposing that
these two factors have in fact tended to offset one another, though the
opposite might be the case in the raw-material countries. For whereas rents,
being largely fixed, rise and fall less than money wages, the price of
imported food-stuffs tends to rise more than money wages in periods of
activity and to fall more in periods of depression. At any rate both Mr
Dunlop and Mr Tarshis claim to show that fluctuations in the terms of trade
(terms of foreign trade in Mr Dunlop's British inquiry and terms of trade
between industry and agriculture in Mr Tarshis's American inquiry) are
not sufficient to affect the general tendency of their results, though
they clearly modify them quantitatively to a considerable extent. Nevertheless, the effect of expenditure on items such as rent, gas, electricity,
water, transport, etc., of which the prices do not change materially in
the short period, needs to be separately calculated before we can be clear.
If it should emerge that it is this factor which explains the results,
the rest of our fundamental generalisations would remain undisturbed. It
is important, therefore, if we are to understand the situation, that the
statisticians should endeavour to calculate wages in terms of the actual
product of the labour in question.
Has the identification of marginal cost with marginal wage cost introduced
a relevant error? In my General Theory of Employment, I have argued that this identification is dangerous in
that it ignores a factor which I have called 'marginal user cost'. It is
unlikely, however, that this can help us in the present context. For marginal
user cost is likely to increase when output is increasing, so that this
factor would work in the opposite direction from that required to explain our
present problem, and would be an additional reason for expecting prices
to rise more than wages. Indeed, one would, on general grounds, expect
marginal total cost to increase more, and not less, than marginal wage
Is it the assumption of increasing marginal real cost in the short period which we ought to suspect? Mr Tarshis finds part of the explanation here; and Dr Kalecki is inclined to infer approximately constant marginal real cost. But there is an important distinction which we have to make. We should all agree that if we start from a level of output very greatly below capacity, so that even the most efficient plant and labour are only partially employed, marginal real cost may be expected to decline with increasing output, or, at the worst, remain constant. But a point must surely come, long before plant and labour are fully employed, when less efficient plant and labour have to be brought into commission or the efficient organisation employed beyond the optimum degree of intensiveness. Even if one concedes that the course of the short-period marginal cost curve is downwards in its early reaches, Mr Kahn's assumption that it eventually turns upwards is, on general common-sense grounds, surely beyond reasonable question; and that this happens, moreover, on a part of the curve which is highly relevant for practical purposes. Certainly it would require more convincing evidence than yet exists to persuade me to give up this presumption.
Nevertheless, it is of great practical importance that the statisticians should endeavour to determine at what level of employment and output the short-period marginal-cost curve for the composite product as a whole begins to turn upward and how sharply it rises after the turning-point has been reached. This knowledge is essential for the interpretation of the trade cycle. It is for this reason that I suggested above that the observations of the relative movement of real and money wages should be separately classified according to the average level of employment which had been reached.
It may prove, indeed, at any rate in the case of statistics relating to recent years that the level of employment has been preponderantly so low that we have been living more often than not on the reaches of the curve before the critical point of upturn has been attained. It should be noticed that Mr Tarshis's American figures relate only to the period from 1932 to 1938, during the whole of which period there has been such intense unemployment in the United States, both of labour and of plant, that it would be quite plausible to suppose that the critical point of the marginal cost curve had never been reached. If this has been the case, it is important that we should know it. But such an experience must not mislead us into supposing that this must necessarily be the case, or into forgetting the sharply different theory which becomes applicable after the turning-point has been reached.
If, indeed, the shape of the marginal-cost curve proves to be such that
we tend to be living, with conditions as they are at present, more often
to the left than to the right of its critical point, the practical case
for a planned expansionist policy is considerably reinforced; for many
to which we must attend after this point has been reached can be, in
that case, frequently neglected. In taking it as my general assumption
that we are often on the right of the critical point, I have been taking
the case in which the practical policy which I have advocated needs the
most careful handling. In particular the warnings given, quite rightly,
by Mr D. H. Robertson of the dangers which may arise when we encourage
or allow the activity of the system to advance too rapidly along the upward
slopes of the marginal-cost curve towards the goal of full employment,
can be more often neglected, for the time being at least, when the assumption
which I have previously admitted as normal and reasonable is abandoned.
There remains the question whether the mistake lies in the approximate identification of marginal cost with price, or rather in the assumption that for output as a whole they bear a more or less proportionate relationship to one another irrespective of the intensity of output. For it may be the case that the practical workings of the laws of imperfect competition in the modern quasi-competitive system are such that, when output increases and money wages rise, prices rise less than in proportion to the increase in marginal money cost. It is scarcely likely, perhaps, that the narrowing gap could be sufficient to prevent a decline in real wages in a phase in which marginal real cost was increasing rapidly. But it might be sufficient to offset the effect on real wages of a modest rise in marginal real cost, and even to dominate the situation in the event of the marginal real cost curve proving to be almost horizontal over a substantial portion of its relevant length.
It is evidently possible that some such factor should exist. It might be, in a sense, merely an extension of the stickiness of prices of which we have already taken account in II above. Apart from those prices which are virtually constant in the short period, there are obviously many others which are, for various reasons, more or less sticky. But this factor would be particularly likely to emerge when output increases, in so far as producers are influenced in their practical price policies and in their exploitation of the opportunities given them by the imperfections of competition, by their long-period average cost, and are less attentive than economists to their short-period marginal cost. Indeed, it is rare for anyone but an economist to suppose that price is predominantly governed by marginal cost. Most business men are surprised by the suggestion that it is a close calculation of short-period marginal cost or of marginal revenue which should dominate their price policies. They maintain that such a policy would rapidly land in bankruptcy anyone who practised it. And if it is true that they are producing more often than not on a scale at which marginal cost is falling with an increase in output, they would clearly be right; for it would be only on rare occasions that they would be collecting anything whatever towards their overhead. It is, beyond doubt, the practical assumption of the producer that his price policy ought to be influenced by the fact that he is normally operating subject to decreasing average cost, even if in the short-period his marginal cost is rising. His effort is to maintain prices when output falls and, when output increases, he may raise them by less than the full amount required to offset higher costs including higher wages. He would admit that this, regarded by him as the reasonable, prudent and far-sighted policy, goes by the board when, at the height of the boom, he is overwhelmed by more orders than he can supply; but even so he is filled with foreboding as to the ultimate consequences of his being forced so far from the right and reasonable policy of fixing his prices by reference to his long-period overhead as well as his current costs. Rightly ordered competition consists, in his opinion, in a proper pressure to secure an adjustment of prices to changes in long-period average cost; and the suggestion that he is becoming a dangerous and anti-social monopolist whenever, by open or tacit agreement with his competitors, he endeavours to prevent prices from hollowing short-period marginal cost, however much this may fall away from long-period average cost, strikes him as disastrous. (It is the failure of the latest phase of the New Deal in the United States, in contrast to the earliest phase, of which the opposite is true, to distinguish between price agreements for maintaining prices in right relation to average long-period cost and those which aim at obtaining a monopolistic profit in excess of average long-period cost which strikes him as particularly unfair.)
Thus, since it is the avowed policy of industrialists to be content with a smaller gross profit per unit of output when output increases than when it declines, it is not unlikely that this policy may be, at least partially, operative. It would be of great interest if the statisticians could show in detail in what way gross profit per unit of output changes in different industries with a changing ratio between actual and capacity output. Such an investigation should distinguish, if possible, between the effect of increasing output on unit-profit and that of higher costs in the shape of higher money wages and other expenses. If it should appear that Increasing output as such has a tendency to decrease unit-profit, it would follow that the policy suggested above is actual as well as professed. If, however, the decline in unit-profit appears to be mainly the result of a tendency of prices to offset higher costs incompletely, irrespective of changes in the level of output, then we have merely an example of the stickiness of prices arising out of the imperfection of competition intrinsic to the market conditions. Unfortunately it is often difficult or impossible to distinguish clearly between the effects of the two influences, since higher money costs and increasing output will generally go together.
A well-known statistical phenomenon which ought to have put me on my guard confirms the probability of constant or diminishing, rather than increasing, profit per unit of output when output increases. I mean the stability of the proportion of the national dividend accruing to labour, irrespective apparently of the level of output as a whole and of the phase of the trade cycle. This is one of the most surprising, yet best-established, facts in the whole range of economic statistics, both for Great Britain and for the United States. The following figures summarise briefly what are, I believe, the undisputed facts:
RELATIVE SHARE OF MANUAL LABOUR IN THE NATIONAL INCOME OF GREAT BRITAIN
RELATIVE SHARE OF MANUAL LABOUR IN THE NATIONAL INCOME OF U.S.A.
The fluctuations in these figures from year to year appear to be of a random character, and certainly give no significant indications of any tendency to move against labour in years of increasing output. It is the stability of the ratio for each country which is chiefly remarkable, and this appears to be a long-run, and not merely a short-period, phenomenon. Moreover, it would be interesting to discover whether the difference between the British and the American ratio is due to a discrepancy in the basis of reckoning adopted in the two sets of statistics or to a significant difference in the degrees of monopoly prevalent in the two countries or to technical conditions.
In any case, these facts do not support the recently prevailing assumptions as to the relative movements of real wages and output, and are inconsistent with the idea of there being any marked tendency to increasing unit-profit with increasing output. Indeed, even in the light of the above considerations, the result remains a bit of a miracle. For even if price policies are such as to cause unit-profit to decrease in the same circumstances as those in which marginal real cost is increasing, why should the two quantities be so related that, regardless of other conditions, the movement of the one almost exactly offsets the movement of the other? I recently offered the problem of explaining this , as Edgeworth would have called it, to the research students at Cambridge. The only solution was offered by Dr Kalecki in the brilliant article which has been published in Econometrica. Dr Kalecki here employs a highly original technique of analysis into the distributional problem between the factors of production in conditions of imperfect competition, which may prove to be an important piece of pioneer work. But the main upshot is what I have indicated above, and Dr Kalecki makes, to the best of my understanding, no definite progress towards explaining why, when there is a change in the ratio of actual to capacity output, the corresponding changes in the degree of the imperfection of competition should so exactly offset other changes. Nor does he explain why the distribution of the product between capital and labour should be stable in the long run, beyond suggestion that changes of one kind always just serve to offset changes of another; yet it is very surprising that on balance there should have been a constant degree of monopoly over the last twenty years or longer. His own explanation is based on the assumptions that marginal real costs are constant, that the degree of the imperfection of the market changes in the opposite direction to output, but that this change is precisely offset by the fact that the prices of basic raw materials (purchased by the system from outside) relatively to money wages increase and decrease with output. Yet there is no obvious reason why these changes should so nearly offset one another; and it would seem safer not to assume that marginal real costs are constant, but to conclude that in actual fact, when output changes, the change in the degree of the imperfection of the market is such as to offset the combined effect of changes in marginal costs and of changes in the prices of materials bought from outside the system relatively to money wages. It may be noticed that Dr Kalecki's argument assumes the existence of an opposite change in the degree of the imperfection of competition (or in the degree in which producers take advantage of it) when output increases from that expected by Mr R. F. Harrod in his study on The Trade Cycle. There Mr Harrod expects an increase; here constancy or a decrease seems to be indicated. Since Mr Harrod gives grounds for his conclusions which are prima facie plausible, this is a further reason for an attempt to put the issue to a more decisive statistical test.
To state the case more exactly, we have five factors which fluctuate in the short period with the level of output:
(1) The price of wage-goods relative to the price of the product;And it appears that, for reasons which are not yet clear, these factors taken in conjunction have no significant influence on the distribution between labour and capital of the income resulting from the output. Whatever a more complete inquiry into the problem may bring forth, it is evident that Mr Dunlop, Mr Tarshis and Dr Kalecki have given us much to think about, and have seriously shaken the fundamental assumptions on which the short-period theory of distribution has been based hitherto;¾ it seems that for practical purposes a different set of simplifications from those adopted hitherto are preferable. Meanwhile I am comforted by the fact that their conclusions tend to confirm the idea that the causes of short-period fluctuation are to be found in changes in the demand for labour, and not in changes in its real-supply price; though I complain a little that I in particular should be criticised for conceding a little to the other view by admitting that, when the changes in effective demand to which I myself attach importance have brought about a change in the level of output, the real-supply price for labour would in fact change in the direction assumed by the theory I am opposing¾as if I was the first to have entertained the fifty-year-old generalisation that, trend eliminated, increasing output is usually associated with a falling real wage.
(2) The price of goods bought from outside the system relatively to money wages;
(3) The marginal wage cost;
(4) The marginal user cost (I attach importance to including this factor because it helps to bridge the discontinuity between an increase of output up to short-period capacity and an increase of output involving an increase beyond the capacity assumed in short-period conditions); and
(5) The degree of the imperfection of competition.
I urge, nevertheless, that we should not be too hasty in our revisions, and that further statistical enquiry is necessary before we have a firm foundation of fact on which to reconstruct our theory of the short period. In particular we need to know:
(i) How the real hourly wage changes in the short period, not merely in relation to the money wage, but in relation to the percentage which actual output bears to capacity output;J. M. KEYNES
(ii) How the purchasing power of the industrial money wage in terms of its own product changes when output changes; and
(iii) How gross profit per unit of output changes (a) when money costs change, and (b) when output changes.