June 6, 2022
C.A.E. Goodhart, Professor Emeritus, London School of Economics
1. What is Money?
Marglin’s magnum opus is so comprehensive, with such a wealth of knowledge on the history of economic thought, that it is a bit churlish to criticise him for not always being right up to date. Yet, as far as I have read, he entirely misses out on those current developments that have put the question of the definition of money back at the forefront of interest. In Chapter 13, money is either commodity money — e.g. silver, cowrie shells, etc. — or bank liabilities, but that is now technologically old hat. It is my view that we cannot begin our discussion of Marglin’s monetary theory on its own terms without first defining the thing being theorized. And that requires grappling, for example, with cryptocurrencies such as Bitcoin, stablecoins such as Facebook’s (Meta’s) Libra (Diem), and the whole range of digital currencies from a variety of issuers, e.g. CBDC.
Perhaps we can be on firmer ground when we describe money as an asset that carries out two main functions: a means of payment and a store of value. But once again, we run into trouble. Bitcoin is a useful means of payment for certain transactions that we want to keep well away from the notice of government, e.g. illegal transactions, but it is a poor medium of exchange for normal transactions and a highly unreliable store of value. And what about Paypal, Alipay, etc.? A U.S. dollar account, in, say, a dollar-backed stablecoin may be a better store of value than a local Latin American currency, but it will be less useful (for small transactions) than local currency. Indeed, an asset can have its capacities to act a means of payment move in an opposite direction from its role as a store of value; witness the simultaneous fall in the use of cash as a means of payment and rise in overall holdings of (large-value) notes during the COVID pandemic.
The money-ness of any asset is not, and cannot be, a 1/0 distinction, with a sharp dividing line between those that are 100% money and those that are 0% money. Marglin does start to enter this issue with his discussion of money market mutual funds (pp. 508-9), but he does not go far enough. I have always had a soft spot for Bill Barnett’s divisia money concept (not mentioned by Marglin), but even that needs reformulation in the light of cryptocurrencies, since the zero interest payable on Bitcoin would make it seem more money-like than it really is.
The production of Bitcoin is (supposed to be) exogenously limited, so it is feasible to imagine a country using only Bitcoin as cash, where interest rates were determined as a spread over that zero interest rate (El Salvador?). But the volatility of Bitcoin, relative to other prices, would make such a financial system so unstable as to be barely useable (see again: El Salvador). Marglin’s breaking out of ‘price risk’ and ‘default risk’ (see below) within the broader range of ‘store of value’ concerns could, perhaps, prove useful in distinguishing money-ness more clearly. Of course, if domestic currency itself became highly unstable, as in Russia during the Ukraine conflict, the disadvantages of Bitcoin would lessen…
2. Marglin’s Main Lesson, and an Extension
I take Marglin’s main conclusion on such monetary matters to be contained in the middle paragraph of the first page of the Conclusions to Chapter 13 (p. 522), as follows:
The homogeneity and substitutability between store of value and medium of exchange make sense in a world of commodity money in which there are no interest-bearing short-term assets. It does not make sense in a world of fractional-reserve banking, in which banks create the bulk of transactions money in response to the public’s demand for the medium of exchange, and in which there are alternative safe assets – safe in the sense both of being free of default risk and of being stable in nominal value. In this world, liquidity preference is a theory of spreads, which requires a monetary authority to anchor these spreads, as central banks have done by fixing the short-term rate.
While I agree with his main point, which is that Keynesian liquidity preference relates to interest rate spreads, not levels, I am disturbed by the fact that virtually every time that he refers to commercial banks in this Chapter he prefixes the reference with the words ‘fractional-reserve,’ as if that feature of banking was the main determinant of their behaviour.
I have always denied this — see Chapter VI on ‘Controlling the Supply of Money: A Problem for the Authorities?’ in Goodhart (Money, Information and Uncertainty, 1975) and several subsequent publications. Nevertheless, the analysis of the determination of the money supply via the money base multiplier, and the focus on the monetary base and the commercial banks’ reserve/deposit ratio, remained mainstream in macro-economics until recently, i.e. serving to emphasise the importance of the fractional reserve aspect of commercial banking. But it will no longer do.
Marglin is absolutely correct when he claims that liquidity preference transforms a theory of spreads into a theory of interest rate levels only in the presence of a zero-interest asset, as shown in the quote below, from p. 514:
An exogenously given money supply, whether commodity money or fiat money, appears to transform a theory of spreads into a theory of levels only because of the addition of a crucial assumption, namely, that wealth portfolios include a zero-interest asset – cash or checking deposits. This assumption allows financial-market equilibrium to determine the levels of various bond rates, including the hurdle rate for investment, because it anchors spreads with an asset yielding a zero rate of interest.”
But what he has failed to recognise is that exactly the same holds true for the mainstream monetary base multiplier, focusing on the fractional reserve base of the banking system. Once authorities offer interest on excess reserves (IOER), as has now become the norm, the previous intimate relationship between changes in central bank interest rates and changes in the reserve base breaks down. Thus he writes earlier (p. 140), “How do we know that banks are fully loaned up? The answer is that we must assume this to be the case, that idle reserves are anathema.” But once IOER is put in place, that assumption collapses. The central bank can, and does, now vary interest rates by announced fiat, without any need for concurrent open market operations. Similarly, the reserve base is determined more by other policy instruments, e.g. Quantitative Easing (QE) or Tightening (QT), rather than open market operations (OMO) and interest-rate changes. The ratio of reserves to deposits has become unanchored and highly variable. The total amount of bank lending and bank deposits is no longer (stably) related to the fractional reserve base, if it ever really was.
Interest rates and monetary growth have become uncoupled, even more than formerly. Nowadays the models that virtually all macro-economists and central bankers use focus entirely on the interest rate transmission mechanism, ignoring completely any direct linkages between monetary growth and nominal incomes. It was not always thus. Although readers may feel that it is something of a digression from Marglin’s arguments, I cannot resist now discussing why this has been so in Section 3 below, and then arguing in Section 4 that continuing to ignore the rate of change of the main monetary aggregates is dangerously short-sighted.
3. Why Central Banks ceased to focus on Monetary Aggregates
Considering how politically dominant Monetarist ideas and policies had become by the end of the 1970s (Bundesbank envy, Thatcher, non-borrowed reserve base, etc., etc.), what is remarkable is the speed with which monetary targetry then disappeared from view towards the end of the 1980s and beginning of the 1990s. There are, perhaps, some four inter-related factors leading to this volte face.
First, the prior econometric evidence on the stability of demand for money functions, and hence for the predictability of the velocity of money, began to break down, particularly for those aggregates chosen by each country to be that country’s monetary target. This observation (as early as 1975) was the basis for ‘Goodhart’s law’.[1] As Governor Bouey of the Bank of Canada remarked, “We did not abandon monetary targets; they abandoned us.”[2] Velocity and demand-for-money functions proved too unpredictable to become a useful intermediate target for policy.
Second, a monetary target was always an intermediate objective for policy. In a world of floating exchange rates, domestic price stability was the ultimate objective for monetary policy. Once that had been codified into the form of an inflation target, the latter could be regarded as a monetary target, adjusted for the wayward and unpredictable fluctuations in velocity. So moving on to an inflation target largely resolved and reconciled the differences between Monetarists and Neo-Keynesians.
Third, central banks had always assessed their own policy position in terms of their ability to control the short-term money market rate, partly because of their historical concern with the external stability of the currency, notably the Gold Standard, and partly because the effects of their policies on such interest rates were immediately visible and clear-cut, whereas the effects (how measured) on the monetary aggregates were far less so. Moreover, the increasingly dominant neo-Keynesian analytical apparatus had the transmission mechanism of monetary policy acting, almost entirely, via the short-term official rate of interest, and expectations of its future values. Practice, and theory, thus focussed solely on this single transmission route, together with expectations of the future paths of such rates and of inflation. In this context what happened to the monetary aggregates could be regarded as an unimportant residual of little relevance.
But most money consists of commercial bank deposits, and banks and their private sector clients, are subject to a variety of forces and influences independent of the official rate of interest. Both Friedman and Keynes were agreed that monetary decisions, for asset purchases and (bank) borrowing as well as deposit holdings, depended on a wide range of relative perceived rates of return, and not just on the margin between bank deposit rates and the official Bank rate. Focussing just on the latter was an extreme form of narrow framing.
But such narrow framing could be, at least partially, excused by our fourth factor, which is that the corridor system of money market operations tended to make (official) interest rate adjustments and (subsequent) monetary aggregate movements correlate (negatively) relatively closely. With a sizeable spread between short-term market rates and the zero interest rate payable on reserves at the central bank, commercial banks’ desired reserve/deposit ratio would (normally) remain stable and predictable, so long as market interest rates remained significantly above zero. Hence the money multiplier, given in the equation M = H (1 + C/D) / (R/D + C/D), would be stable, and open market sales by the central bank aimed at raising the official short rate could also be reasonably confidently assumed to reduce bank borrowing and bank deposits. So there was little or no extra information (about other factors influencing monetary decisions) to be gained by looking at the monetary aggregates, and anyhow velocity was unpredictable (see the first and second factors above).
4. Have the Monetary Aggregates become, once again, useful Information Variables?
The adoption of a floor system of money market operations in 2008 and the fall of interest rates to the zero lower bound (ZLB), plus QE, have completely undermined the stability of the R/D ratio and hence of the money multiplier. The forces tending to cause the monetary aggregates and official interest rates to move negatively in tandem have weakened sharply. Thus the monetary aggregates now have greater freedom to respond to the myriad other factors driving monetary decisions. To this extent the monetary aggregates may now play a greater role than in the decades up to 2008 as information variables, reflecting forces beyond the narrow framing of the standard three-equation Neo-Keynesian model.
The problem, however, is that velocity remains quite unpredictable, an unpredictability that would only worsen were we to use the aggregates as control variables. So, is a sizeable move in a monetary aggregate a signal providing useful information on wider macro-economic conditions, or just an erratic shift in velocity? Some economists are predisposed towards the former view; others to the latter. Given the present limited state of macro-economic analysis, it has to be a matter of judgement. But when recent fluctuations in monetary growth have been so large, to dismiss these, tout court, as just a meaningless shift in velocity strikes me as dangerously complacent.
[1] “Any observed statistical regularity will tend to collapse once pressure is placed upon it for control purposes.” Charles Goodhart, “Problems of Monetary Management: The U.K. Experience,” in Inflation, Depression, and Economic Policy in the West, ed. Anthony S. Courakis (1975), 116. BACK TO POST
[2]Canada, House of Commons Standing Committee on Finance, Trade and Economic Affairs, Minutes of Proceedings and Evidence, No 134, March 28, 1983, p. 12. BACK TO POST