Fall 2021 - A Symposium on: Stephen A. Marglin’s Raising Keynes: A Twenty-First Century Theory
Michael Woodford, Is the Demand for Money the Source of Economic Slumps?

Print Friendly, PDF & Email
Forms of Money
Forms of Money by James Rose


March 25, 2022

Michael Woodford, Columbia University 

This piece, focusing on the monetary theory topics raised in the thirteenth chapter of Stephen Marglin’s Raising Keynes, is adapted from a longer essay discussing other aspects of the analysis. The full essay can be found on my website.
The key insight of Keynes’ General Theory is sometimes said to have been the recognition that actual economies are monetary economies rather than barter economies — that is, that goods and services are sold for money rather than directly exchanged for other goods and services. This is thought to be the key to understanding how a situation of persistently insufficient aggregate demand is possible. In a (hypothetical) barter economy, it is argued, an offer to supply goods or services would have to be simultaneously an offer to buy other goods and services of equal value from someone else, so that there would be no possibility of an overall shortfall of demand relative to supply.
An influential formulation of this idea during the heyday of general equilibrium theory was due to Don Patinkin. The key puzzle, for Patinkin, was how an insufficiency of aggregate demand could be consistent with Walras’ Law: the proposition that optimizing supplies and demands should satisfy the property that, when the supplies and demands for different goods are aggregated using market prices, the total excess of supply over demand equals exactly zero. This property should hold regardless of the prices that economic units take as given in choosing the amounts that they wish to supply or demand; hence an assumption that wages or prices are rigid for some reason at levels that do not clear markets is not a reason for an exception to Walras’ Law. Patinkin’s solution was to argue that in a monetary economy, money balances had to be included as one of the “goods” demanded by economic units; the Keynesian scenario of a slump due to a shortfall of aggregate demand was then imagined to involve demand less than supply in the case of goods other than money, but an excess demand for money.
Under this view, one might suppose that study of the sources of the demand for money balances would be a crucial issue for macroeconomic theory. One might also imagine that innovations in the payment system that affect the supply of and demand for money-like assets should have fundamental consequences for the overall level of business activity. Assigning such a crucial role to money demand has been, I believe, a fundamental misunderstanding, and one that has resulted in much wasted effort on the part of macroeconomic theorists during the second half of the twentieth century.
One of the great services of Stephen Marglin’s masterful restatement of the argument of the General Theory is to make it clear that the basic logic of a Keynesian slump can be understood without any reference at all to a demand for money balances. It is true, of course, that Keynes discusses the demand for money in the General Theory, and indeed his discussion of the “liquidity preference function” is often considered one of the book’s most important theoretical innovations. Marglin discusses this aspect of Keynes’ theory, in considerable detail but it is one of the respects in which he is most critical of Keynes’ formulations.
The central problem for Keynes — and the central problem addressed by Marglin in Raising Keynes as well — is explaining how a sustained economic slump like the Great Depression is possible; in particular, explaining why there are no automatic market mechanisms that should eliminate an imbalance between the aggregate demand for produced goods and services and the quantity that firms would wish to supply at current prices and wages. A view that Keynes was especially concerned to refute was the doctrine, common among his contemporaries, that interest rates should adjust so as to ensure balance between current desired spending and desired supply. (This was often expressed by saying that interest rates should bring investment into line with savings, but this is equivalent to saying that they should bring aggregate spending into line with aggregate supply.)
Keynes refuted this doctrine by offering an alternative account of how interest rates are determined: his famous “liquidity preference theory” of interest rates. According this proposal, interest rates adjust so as to cause the private sector to willingly hold money balances of exactly the quantity that exists. Specifically, if desired holdings of real money balances are an increasing function of the volume of real economic activity (constituting “transactions demand”) and a decreasing function of the nominal interest rate (the opportunity cost of holding non-interest-earning money balances), then one obtains a relationship between the nominal interest rate and the level of real activity that must hold in order for desired real balances to equal the real purchasing power of the existing money supply. (This is the “LM curve” of John Hicks’ IS-LM diagram, a staple of undergraduate textbook expositions of Keynesian economics.)  
If one combines this with a relation (the “IS curve”) that expresses desired current aggregate spending as a function of the interest rate and other factors such as expectations regarding future income, one can solve for equilibrium values of both output and the interest rate as functions of the real money supply and expectations. This determines an equilibrium level of aggregate demand, that is unrelated to the factors (such as attitudes toward labor supply and firms’ production technologies) that determine aggregate supply. Hence there is no reason in general that aggregate demand should be consistent with goods market clearing, if we take the level of prices as given.
Marglin objects to this theory, because while it succeeds in explaining why aggregate demand need not coincide with aggregate supply at every instant, it also provides a mechanism through which aggregate demand shortfalls might be expected to be automatically corrected. The discussion above assumes as given both the nominal money supply and the general level of prices, which determines the purchasing power of that money supply. But if aggregate demand falls short of the amount that firms would wish to supply at current prices, firms have a reason to cut their prices. And if one assumes the nominal money supply to remain unchanged (at its initially given value), falling prices will mean an increase in the real money supply. This in turn makes a larger volume of real transactions consistent with equality between the real money supply and desired real balances unless interest rates decline; but lower interest rates should imply an increase in desired current spending so that some increase in the volume of real transactions should occur. Since prices should continue to fall as long as aggregate demand remains insufficient to support the Walrasian equilibrium, this process should continue until the shortfall of aggregate demand is eliminated. (This mechanism is collectively known as “the Keynes effect.”)
Keynes’ response to this possibility was to argue that when interest rates become very low, the demand function for real money balances becomes infinitely elastic. This means that no further change in either the volume of transactions or interest rates will be needed in order to get the private sector to hold larger real money balances.  In the case of such a “liquidity trap,” further decreases in prices would no longer increase aggregate demand. Keynes argued on this ground that a slump with unemployed resources could be sustained indefinitely as an equilibrium under circumstances where aggregate demand remains low, even in the case of an interest rate that has fallen to its lower bound. Yet many critics of Keynesian economics questioned how common such a situation would be, if it could occur at all. Others argued that increases in the real money supply should increase aggregate demand through other channels (such as the “real balance effect” hypothesized by Arthur Pigou and Don Patinkin, among others), even if the demand for liquid balances were to become infinitely elastic.
Marglin instead protests Keynes’ assumption of a fixed money supply, arguing that this might be reasonable as a model of a commodity-money system (in which the money supply might be determined by the existing quantity of some real commodity, such as gold), but that it is grossly inadequate as a representation of a modern monetary system. He argues that the relevant monetary aggregate for Keynes’ “liquidity preference function” should include the money-like liabilities of private institutions (such as transactions balances held at commercial banks), and that the quantity of such liabilities should vary endogenously with business conditions. In particular, he notes that in crises like the Great Depression, the supply of privately-created money often contracts sharply, as a result of loss of confidence in the issuing institutions. When this happens, falling prices need not imply an increase in the real money supply — as indeed they did not during the Great Contraction of 1929-33 in the U.S., despite severe price deflation.
Marglin also criticizes Keynes’ liquidity preference theory of interest rates in a more fundamental way, pointing out that the opportunity cost of holding money that appears as an argument of the “liquidity preference function” should really be an interest-rate differential. That is, it should be the difference between the interest rate available on non-monetary stores of value and that available on money balances; this only becomes equal to the nominal interest rate under the assumption that the interest rate paid on money is necessarily zero. Thus Marglin argues that even on the assumption of a given money supply, one cannot have a complete theory of interest-rate determination that is independent of government policy. 
The equilibrium nominal interest rate on non-monetary assets should depend on the interest paid on money balances, and this ultimately depends on government policy. In particular, it depends on the interest rate that the central bank pays on overnight balances held with it, and more generally on the complex of interest rates associated with various standing facilities operated by the central bank under a typical modern monetary system. Thus, it makes no sense to speak of the level of interest rates as being “market determined,” even on the assumption of a fixed money supply. Interest rates and aggregate demand are both strongly influenced by decisions of the central bank, not only through the ability of central banks to control the quantity of money (emphasized by mid-20th century monetarists), but also through central bank control of various administered interest rates (such as the interest rate on reserve balances, or IORB, a central policy instrument of the Federal Reserve under current operating procedures).
These criticisms are well-taken, though they are not entirely novel. Marglin’s view of the nature of money is actually quite consistent with much of the current mainstream literature. The New Keynesian literature that has developed since the 1990s — since the Federal Reserve’s shift adoption of a policy of publicly announcing its operating target for the federal funds rate (rather than remaining opaque about whether it controlled short-term interest rates) on the one hand, and following the rise to popularity of formulations like the “Taylor rule” as simple descriptions of actual central-bank policies — has emphasized central-bank control of a short-term nominal interest rate as a key determinant of aggregate demand. Discussions of how the central bank’s target for the policy rate is implemented in practice stress the relevance of both quantity adjustments (variations in the supply of bank reserves through open-market operations) and adjustments of administered interest rates. Theoretical expositions of how this works often use an equilibrium relation in which desired real money balances are a function of an interest-rate differential rather than the absolute level of some nominal interest rate, just as Marglin proposes. (See for example the treatment of these issues in chapters 1 and 2 of my book.)
It should also be noted that while Marglin criticizes Keynes for relying upon the doctrine of the “liquidity trap” to explain how a sustained slump is possible, the Keynesian liquidity trap has proven to be a surprisingly relevant idea for understanding macroeconomic developments and policy challenges in the wake of the global financial crisis of 2008. It is true that the understanding of the liquidity trap in the modern literature is somewhat different than the argument presented in the General Theory. Keynes argued that the demand for money should become highly elastic in the case of a low enough yield on long bonds, as a result of speculation about the future value of such bonds; the modern literature instead stresses that the demand for base money becomes highly elastic when the differential between other money-market interest rates and the interest paid on central-bank balances becomes negligible. But this latter insight has proven to be correct, and crucial for understanding why massive expansions of central banks’ monetary liabilities in response to the crisis were not inflationary (contrary to the warnings of many commentators who had never accepted the logic of the General Theory). 
And the simple idea of a lower bound, not simply on the interest differential, but on the absolute level of money-market interest rates, turned out to be right. While the form of the liquidity preference function only limits the ability to reduce the interest differential, there have also proven to be important practical limits on the ability of central banks to reduce the interest rate that they pay on balances held at the central bank much below zero.[1] Thus, the fact that interest rates can be influenced by varying the interest paid on central-bank balances is important when thinking about the ways in which a central bank can raise interest rates well above zero; it implies, for example, that restraint of aggregate demand for the sake of inflation control need not require a central bank to reduce the size of its credit or liquidity programs, in order to be able to contract the size of its balance sheet. But this refinement of traditional theory does not much change one’s view of policy options in a deflationary crisis, since in such a situation the central bank will wish to maintain the interest rate on central-bank balances fixed at a lower bound that is near zero, even if it recognizes that this interest rate remains an additional policy instrument.
Thus Keynes’ idea that there should be a bound on how far nominal interest rates can fall, even in the case of a massive expansion of the monetary liabilities of the central bank, has been tested by many of the leading central banks (both in response to the crisis of 2008 and in response to the COVID-19 pandemic), and found to be true, at least to a first approximation. This has led policymakers to contemplate other tools through which to influence aggregate demand — renewed interest in fiscal stimulus policies, novel experiments with forward guidance in monetary policy, and more direct involvement in the provision of credit to non-financial institutions, among others.
Nonetheless, I agree with Marglin that Keynes’ discussion of the role of money in aggregate demand determination is one of the least satisfactory parts of the General Theory — and in particular, one of the aspects of Keynes’ theory that is least relevant to contemporary circumstances. I also agree that this is not really essential for understanding how a sustained slump is possible.
The key to Keynes’ refutation of the classical doctrine, according to which interest rates must adjust to ensure a level of aggregate demand in line with aggregate supply, is not the ideas summarized by the “LM curve,” but rather those summarized by the “IS curve.” Each of the points on the latter curve represents a situation in which there is equilibrium in financial markets, but there is not just one real interest rate consistent with this curve — instead, a continuum of different real interest rates are possible, each associated with a different level of current aggregate demand. The requirement of financial market equilibrium does not determine which of these points the economy will operate at. Instead, central-bank interest-rate policy determines a particular level of interest rates and, associated with them, a particular level of aggregate demand. (The “LM curve” represents one way of understanding how the central bank is able to influence interest rates, but this rather old-fashioned account of how central banks implement their interest-rate policies is not needed, for the reasons that Marglin discusses.)
It is indeed important to the Keynesian analysis that actual economies use money rather than trading goods directly for other goods (or services such as labor), but the reason does not have to do with the nature of the demand for money balances. What is important is the role of money as a unit of account, rather than as a potential store of value. It is crucial to Keynes’ discussion (and Marglin’s exposition of it) that wages and prices are quoted in monetary units. This means that when wages and/or prices are perceived to be out of line with current market conditions, they are adjusted in monetary terms. This in turn makes possible a continuing slump, in which both wages and prices are constantly falling in monetary terms in response to the insufficiency of demand, but relative prices (such as the real wage) do not change. Nor need there be any change in the perceived relative price of current goods and services in terms of future goods; if people expect a particular rate of inflation (as opposed to having definite expectations about the future price level), then decreases in current prices and wages in money terms automatically imply correspondingly lower expected future prices and wages in monetary terms as well, so that there is no change in perceived intertemporal relative prices.
The fact that cutting prices and wages in monetary units need not make any difference for relative prices is a crucial observation. Because of this, the mere fact that households and firms should both be motivated to cut the prices and wages that they demand in the case of a continuing slump does not mean that there are automatic market forces that must move the economy to a Walrasian equilibrium. Marglin shows in great detail how this is possible — and indeed, that such an imbalance (of one sign or the other) is inevitable in a monetary economy in the absence of a suitable policy response. (Elsewhere I further discuss aspects of Marglin’s analysis, and show how micro-founded New Keynesian models have similar implications.)
Thus it is essential to recognize that actual economies are monetary economies, rather than imagining (as in typical expositions of general equilibrium theory) that market mechanisms determine relative prices without any need for reference to a monetary unit of account. But, this does not mean that the central challenge for macroeconomic theory should be to better understand the nature of money demand, nor that monetary innovation is likely to be the key to increased macroeconomic stability. Instead, better understanding of the adjustment dynamics of nominal wages and prices is what would matter most for improving the models used for policy analysis. And the kind of potential innovations that are most likely to change the way the economy functions would be ones that change the terms in which households and firms negotiate prices and wages.

[1] Bernanke discusses the scope of negative rates in practice and the practical obstacles to their use at his blog. BACK TO POST


Return to Symposium Prompt