TL;DR: “Supply creates its own demand” >> “Production must precede exchange”. Most anti-Keynesians get this completely backward, because they fail to realize that the ex ante/ex post distinction is the key to understanding the literature on Say’s Law.
To elaborate a bit on the above: this question is important because there is an unfortunate tendency among anti-Keynesians to offer “production must precede exchange” as a would-be correction to Keynes’s “supply creates its own demand”, on the assumption that this latter construct has the absurd implication that Say’s Law views recession and unemployment as impossible.
But any consideration whose logic reduces to “supply is a necessary condition for demand” is ultimately worthless in the debate over the foundations of macroeconomics – this proposition is both (mostly) untrue and irrelevant. It is irrelevant because even if taken at face value, it does not even attempt to deny that demand can fall short of supply (due to deficient willingness to demand), which is really the entire point at issue. And as a result, it is not even substantially true – supply is only a necessary condition for demand in the vacuous sense that anything consumed must have been produced at some point in time. But if we grant the possibility that there may have been a shortfall of demand relative to supply in some prior instance, then in the present moment, we do not necessarily need to supply in order to demand. As any good Keynesian will tell you, we merely need to put purchasing power in the hands of people who are willing to use it.
So if Keynes’s “supply is a sufficient condition for demand” is absurd, and “supply is a necessary condition for demand” is uninteresting, where does that leave us? The answer is that “supply is a sufficient condition for demand” is only problematic if we insist on interpreting it in an ex post sense – as a claim that everything produced actually ends up being demanded. But in fact, when Keynes gives his technical definition of Say’s Law, he is quite explicit that it is to be understood ex ante. (He defines the “aggregate demand price of output” as “the amount that entrepreneurs EXPECT to receive” – my emphasis. It is hard to get much more ex ante than that). So Keynes’s definition of Say’s Law (which he rejects as being false) can be summarized as “supply is a sufficient condition for ex ante demand”. This is a perfectly viable and interesting proposition, and in my opinion, it is clearly the very same proposition that was affirmed by the Classical economists, most explicitly by J.S. Mill. So despite all the screeching to the contrary in the anti-Keynesian literature, Keynes did not misunderstand Say’s Law at all.* Where he went wrong was in rejecting it for completely insufficient reasons.
In my opinion, much of the fog that continues to surround Say’s Law discussions is caused by the logical ambiguity of the notion that supply constitutes demand – the point that Say himself makes repeatedly in his Traite. If you take this consideration on its own, it seems designed to support the conclusion that supply is a necessary condition for demand. But in Mill, “supply constitutes demand” is invoked in a different logical context. Mill’s reasoning goes as follows: 1) Ability to purchase and willingness to purchase, taken together, are sufficient conditions for (ex ante) demand. 2) When a producer creates a given increment of supply, he “shows [himself] to have the desire, and obtains the means, of consuming” – this can be summarized as supply constitutes demand, and supply also implies demand. Therefore 3) Supply is a sufficient condition for ex ante demand. Here, “supply constitutes demand” has nothing to do with supply being a necessary condition for demand; rather, it shows that supply satisfies one of the two conditions for demand that are sufficient when taken together.
But is this conclusion really “Say’s Law”, or did Mill just modify Say’s reasoning as he saw fit? Well, for one thing, even if he did, he was clearly justified in doing so, as we saw above that “supply is a necessary condition for demand” is a non-starter. But personally, I don’t think that Say ever intended to suggest this inferior version – for one thing, in his chapter in his Principles book, after explicitly making a “supply is a sufficient condition for demand” argument, Mill credits Say (along with Ricardo) with having reached the same conclusion. Also, even if Say is disappointingly unclear on this in the Traite, it is pretty clear from his letters to Malthus that he intended to claim that demand deficiency was an absolute impossibility, a la Mill. My opinion is that Say had much the same logical structure in mind as Mill, but neglected to pay much attention to the “willingness to demand” plank because he thought it completely obvious that this element could not be lacking. (Recall that when he wrote the Traite, Malthus had not yet made this point the focus of his objections, and that Say’s discussion was primarily aimed at the “mercantilist” failure to understand that production, not money per se, is what ultimately generates purchasing power).
To sum up, here are my theses about Say’s Law. The more I have studied/thought about this issue over the past year, the more I am convinced that they are correct.
1) Say’s Law is best defined as “supply is a sufficient condition for ex ante demand”. This is the proposition that was affirmed by Mill, Say, and Ricardo, and denied by Keynes.
2) Keynes did not fundamentally misunderstand Say’s Law. There is genuine logical clash between his position and Mill’s.
3) After Keynes, virtually all discussion of Say’s Law has completely missed the point. Virtually without exception, it has consisted of either:
a) The claim that Say’s Law denied the possibility of recession/unemployment
b) The claim that Keynes was responsible for the misunderstanding in a), often accompanied by the purported correction that Say’s Law actually says that production must precede exchange (or something similar)
c) The mainstream definition of Say’s Law as an assumption of monetary equilibrium, given first by Oskar Lange and later popularized by Becker and Baumol as Say’s Identity and Say’s Equality.
If there is any discussion out there that avoids all of these (as I view them) pitfalls, I would love to hear about it!
* Actually, this is not quite true – there IS a subsidiary sense in which Keynes appears to have misunderstood the Classical discussion. But he is absolutely innocent of the charges leveled at him by the likes of Hazlitt, Hutt and Kates.