The General Theory and the atomic hypothesis
Discerning the role of the atomic hypothesis in GT is, I think, not a simple matter but a rather delicate and somewhat intricate one. It seems quite different from the example of economic men's conventional behaviour in the stock exchange identified in the Quarterly Journal of Economics article.
First of all, here is Keynes's definitive assertion about the main topic that we intend to consider:
I have called my theory a general theory. I mean by this that I am chiefly concerned with the behaviour of the economic system as a whole,—with aggregate incomes, aggregate profits, aggregate output, aggregate employment, aggregate investment, aggregate saving.... And I argue that important mistakes have been made through extending to the system as a whole conclusions which have been correctly arrived at in respect of a part of it taken in isolation.
(GT, p. xxxii, Preface to the French Edition , emphasis added)
Though an individual whose transactions are small in relation to the market can safely neglect the fact that demand is not a one-sided transaction, it makes nonsense to neglect it when we come to aggregate demand. This is the vital difference between the theory of economic behaviour of the aggregate and the theory of the individual unit, in which we assume that changes in the individual's own demand do not affect his income.
(GT, p. 85; emphasis added)
A typical example of this 'difference between the theory of economic behaviour of the aggregate and the theory of the individual unit' is (though it might be needless to mention) the fallacy of composition shown by the so-called savings paradox. Any individual is, whatever his own motive of saving, able to increase his wealth by increasing his savings. This is, undoubtedly, a 'correct proposition' to him. But we cannot maintain that his saving behaviour contributes to increasing pro tanto aggregate wealth as well through extending to the system as a whole such a correct conclusion. Actually, such reasoning falls within the category of the 'fallacy of composition'. In this respect, a reduction in the individual's consumption associated with his or her saving behaviour affects the incomes of other people as well, so that we may expect changes in the aggregate income through such a series of chain reactions. In short, a manner of reasoning aiming at general proposition valid in the whole through the adding up of micro-level conclusions valid in a part of it obviously overlooks such complicated chain reactions ultimately affecting aggregate income.
Although 'the amount of saving is an outcome of the collective behaviour of individual consumers' (GT, p. 63), the existence of the savings paradox is absolutely not compatible with methodological individualism, the crux of which is to construct general laws or uniformity in the system as a whole through reducing it into the separate distinct and rational behaviours of individual units and then basically relying on normal individual behaviour.
Another typical example of the 'difference between the theory of economic behaviour of the aggregate and the theory of the individual unit' is the demand schedule for labour. Since Keynes's world explicitly accepts the 'first classical fundamental postulate', he is obliged to admit the inverse functional relation of employment to wages in particular industries. Thus, Keynes states that 'no one would wish to deny the proposition that a reduction in money-wages accompanied by the same aggregate effective demand as before will be associated with an increase in employment' (GT, p. 259; emphasis in original).
It is precisely a fallacy of composition that one endeavours to obtain 'a demand schedule for labour in industry as a whole' through extending by analogy such conclusions in respect of a particular industry and to find therein the inverse functional relation as well.
The reason such an analogy is wrong is that, according to Keynes, it entirely overlooks the effects of chain reactions starting from a particular industry and ultimately changing aggregate effective demand, just as in the savings paradox.
Since we do not mean to admit (though we do not discuss it further) the 'first classical fundamental postulate', that is, that 'the wage is equal to the marginal product of labour', we cannot regard 'the demand schedule for labour in the industry relating the quantity of employment to different levels of wages' (GT, p. 259) as a consistent and solid ground for economic theory.9
However, when Keynes comes to the cost side his approach suddenly and fundamentally changes. A suitable approach on the cost side must be 'to extend by analogy its conclusions in respect of a particular industry to industry as a whole' (GT, p. 260). It is worth recalling that this is precisely the method rejected by Keynes on the demand side. In rough terms, since the behaviour of an individual part and the behaviour of the whole are, on the cost side, not basically distinct from each other (except for the degree of aggregation), one is liable to adopt the method by which one first disaggregates the whole into separate parts and then extends by analogy the conclusions arrived at in respect of a particular component part to the system as a whole. This is, I think, nothing other than methodological individualism, or the atomic hypothesis.
In any case, it is interesting to note what Keynes asserts:
In a single industry its particular price-level depends partly on the rate of remuneration of the factors of production which enter into its marginal cost, and partly on the scale of output. There is no reason to modify this conclusion when we pass to industry as a whole. The general price-level depends partly on the rate of remuneration of the factors of production which enter into marginal cost and partly on the scale of output as a whole, i.e. (taking equipment and technique as given) on the volume of employment.
(GT, p. 294; emphasis added)
As seen above, it is obvious that the Marshallian short-run and diminishing returns (that is, increasing costs) industry are the starting point of Keynes's cost analysis. What is rather peculiar to Keynes's approach and makes it somewhat different from Marshall's is that it extends to industry as a whole the positive monotonic relation of marginal cost to output in respect of a single industry without any modification. As a result, Keynes's production economy (viewed from the 'cost' angle) seems to be something like a gigantic industry composed of homogenous and continuous elements.10 It forms a conspicuous contrast with the analysis of Piero Sraffa, in which the economic system generates the objective technological network among various industries, and the general price-level associated with industry as a whole must be represented explicitly as a set of particular prices. But, according to the assumption of increasing costs, the Keynesian marginal cost of such an economy-wide industry, if it were to be, is assumed to rise as the scale of output as a whole increases.
It is true that, since Keynes maintains that 'when we pass to output as a whole, the costs of production in any industry partly depend on the output of other industries' (GT, p. 294), he would seem to take into account the technological network among industries. But, in fact, this remains on the fringe of his attention. This is shown by the fact that, starting from the short-run diminishing returns of a particular industry considered in isolation, Keynes extends by analogy the conclusions attained in respect of a single industry (increasing marginal cost) to the output as a whole. This is, I think, unquestionable.
Thus Keynes, within the above conceptual framework, insists that the importance of his remarkable and revolutionary contribution lies on the demand side of economic theory. And he states:
It is on the side of demand that we have to introduce quite new ideas when we are dealing with demand as a whole and no longer with the demand for a single product taken in isolation, with the demand as a whole assumed to be unchanged.
(GT, p. 247; emphasis added)
Be that as it may, it follows from what we saw above, first, that Keynes explicitly grounds the atomic hypothesis on the side of cost and, second, that he conceives as the starting point of his cost analysis the Marshallian short-run supply curve for a single product taken in isolation.11