Whatever became of the Monetary Aggregates?

Whatever became of the Monetary Aggregates?

As adapted from the Peston Lecture in honour of Maurice, Lord Peston,

delivered at Queen Mary College, London, on February 28, 2007

By Charles Goodhart

My title intentionally harks back to Maurice Peston’s slim, but excellent, 1980 book,

entitled Whatever happened to Macro-economics. In this book, a compilation of three

lectures, Maurice asks how much then remained of traditional Keynesian macroeconomics

in the aftermath of the Monetarist counter-revolution, and of the

development of Lucasian rational expectations. Maurice was much more impressed

by the new contributions of the rational expectations school than he was of those by

the more traditional Monetarists.

Indeed, time has appeared to prove Maurice to be correct in this appreciation. After

all, the monetary aggregates, the money supply in one, or other, of its various guises,

should presumably play a major role in any monetarist scheme of affairs. As Mike

Woodford has recorded, in a recent paper,

“[N]owadays monetary aggregates play little role in monetary policy

deliberations at most central banks.”

In contrast, a detailed treatment of expectations, and how these may be generated, lies

at the heart of the current Neo-Keynesian analysis.

2

My own thesis is that this downgrading of the role of the monetary aggregates in

current models, and in forecasting future inflation, has gone too far. At this point a

couple of personal caveats may be in order. I am far from being a card-carrying

monetarist. Not only did I strenuously oppose Friedman’s monetary base control

mechanism and his K% rule for monetary growth, but I have been credited, though

without much justification, for example in The Times obituary of Milton Friedman,

(November 17, 2006), with having undetermined monetarism by pointing out the

likely instability of demand for money functions when turned into targets. That said, I

was shocked when successive Conservative governments in the 1980s and 1990s

could pass almost seamlessly from the view that control over broad money was the

essential centrepiece of macro policy, the Medium Term Financial Strategy, to

subsequently paying little, or no attention to monetary developments in later years.

By the same token I am concerned that some of the key features of a monetary, and of

a truly Keynesian, economy are ignored in the neo-classical (neo-Keynesian)

consensus.

A second caveat is that Woodford, and his supporters at the recent ECB Conference in

Frankfurt (9/10 November, 2006) on ‘The Role of Money’, such as Uhlig and Gali

had, I believe, two separate purposes. The first was to deny the benefit of having a

separate monetary analysis, the famous two pillars of the ECB. In so far as monetary

effects were important in preparing forecasts, and deciding policy, they should be

integrated into a single, overall analysis of the prospects for the economy. I have no

quarrel with this. It is the further second line of argument, that we can in practice

virtually ignore developments in the monetary aggregates in the conduct of monetary

3

policy, focussing instead solely on a key policy determined interest rate, that I do

want to question.

The starting point for Woodford (2006), as also for a somewhat similar, earlier article

by B. McCallum, (2001), ‘Monetary Policy Analysis in Models without Money’, is

the basic, now widely accepted, new Keynesian model of three equations, consisting

of an IS type aggregate expenditure function, a Phillips-curve type aggregate supply

function, and a Taylor-type Central Bank reaction function, showing how Central

Banks set interest rates.

In both the IS and the Phillips supply curve, expectations play a leading role; and

whether they are forwards or backwards looking, rational or bounded, forms the core

of a huge literature, but one which is not relevant to my thesis here. So I shall more

simply write these as:-

ỹ

= yt – y* = f(Eỹ, R – Eπ) + ut 1 (IS)

π

where

sustainable level of output, R is the nominal interest rate, E is the expectations

operator,

This is complemented by the Taylor type reaction function:-

R

Where

4

Let me make two peripheral brief comments on this. First, it seems odd that the

private sector is shown as responding to future expectations, but in equation 3 the

Central Bank appears to be reacting only to current events. This latter is surely

wrong. All inflation targeting Central Banks make, and respond to, inflation

forecasts. I have tried elsewhere to show that a proper forward-looking specification

of reaction functions can make a large difference to the estimated coefficients,

(Goodhart, 2005).

Second, this three equation model may only work satisfactorily during relatively calm

periods of stable economic developments, a ‘fair weather’ model. The zero bound to

nominal interest rates suggests that this model may have limited usefulness during

periods of deflationary pressures. Moreover, during turbulent periods, whether of

severe deflation or inflation, expectations will not be anchored, will differ quite

markedly from person to person, and be subject to potentially rapid and sharp

revision. Under these circumstances one cannot assess what real rates of interest may

be, so growth rates of the monetary aggregates may well be a better guide to the

effects of monetary policy on the economy than either nominal, or an estimate of real,

interest rates.

Be that as it may, this three equation model determines the level of interest rates, (real

and nominal), the output gap, and both inflation and the price level, (given the

inflation target and the starting point). The system has a determinate equilibrium, so

long as the Central Bank reacts sufficiently strongly to inflation in adjusting nominal

interest rates. There appears to be no need to look at what is happening to money in

5

this system to achieve the main macro-economic variables of importance to social

welfare.

In practice, however, money can be, and generally is, present in this model, since a

demand-for-money function is fully consistent with the above three equations, as in

equation 4:-

M

where P

the Central Bank sets interest rates, as is the generality, the money stock is a

dependent, endogenous variable. This is exactly what the heterodox, Post-

Keynesians, from Kaldor, through Vicky Chick, and on through Basil Moore and

Randy Wray, have been correctly claiming for decades, and I have been in their party

on this. Certainly if the demand for money function fits perfectly, and if its

arguments are correctly set out in equation (4), then you learn nothing more from

looking at money, than you already knew from looking at inflation, output and

interest rates.

There is a minor caveat to this, which is that money stock data, or elements of it, may

come out earlier, or be less subject to (initial) measurement errors than data on output.

If so, M could act as an early indicator variable for y, or even for Py. There have been

a few instances of such indicator relationships; for example fluctuations in cash

holdings (and M0), at least for a time, seemed to precede movements in personal

consumption in the UK. I would not, myself however, want to put much weight on

any such leading indicator properties.

6

But much the same is true of output, or of inflation; they too are endogenous,

dependent variables. If their functional relationships held perfectly, then you would

learn nothing more from looking at output, or inflation, outcomes than one already

knew from a knowledge of its functional (structural) relationship. In reality, however,

a great deal of time and effort is spent analysing whether the deviation of

output/inflation from forecast is due to one kind of shock, or another, e.g. transient or

permanent, demand or supply shock. Why do we not care as much to shocks, to

deviations of the money stock, from its expected value?

One great advantage of Central Banks running monetary policy by setting interest

rates, rather than via the monetary base, is that demand-side shocks to desired money

holdings are automatically accommodated. So if one should assume that all monetary

shocks are demand-side, then movements in the money supply indeed tell one

nothing. Note, however, that demand-side shocks to the real economy are also

relatively easy to handle, in theory at least.

But it is not true that all shocks to money are demand-side. The bulk of money is in

the form of commercial bank liabilities, and banks can behave very differently over

time. The form of their liabilities , their capital base, their confidence, their risk

appetite can and does alter over time, both cyclically and more permanently. The

whole question of whether certain segments of the economy can access funds beyond

their current income depends crucially on the behaviour of the banks. If there is a

supply shock to money, with certain groups now getting more, or less, access to

funding, for example when banks provide mortgages to a wider group of households

7

on easier terms, will this not feed back into the IS curve? Of course it will. I shall

revert to this in a deeper format later on.

But, first, I want to turn to two other issues. The first of these relates to the much

closer long-run, low frequency, relationship between money and prices than is

observed at higher frequency, shorter run periods, when the relationship is obscured

by a variety of shocks. Michael Woodford accepts in his Frankfurt paper, slightly

grudgingly for my taste, this longer term relationship, but argues that it is an inherent

consequence of having a reasonably stable long-run demand-for-money function.

Thus, if we first difference equation (4),

t = f(Eπ, ỹ) + vt 2 (AS)ỹ is the output gap, y is current real output, y* is the natural, or equilibrium, orπ is the rate of inflation, and u and v are error terms.t = a + b1(πt – π*) + b2ỹ (3)π* is the target inflation rate.t/Pt = f(y, R, π) + wt (4)t is the price level and wt another error term. Note, however, that so long as

μ

(see his equation 3.3), it can be easily shown “that every term on the right-hand side

of (3.3) is stationary, so that

of the money stock and of prices must, in the medium and longer run, move in

tandem.

That is surely correct, but Woodford then goes on to argue that looking at inflationary

outcomes by themselves is just as good, or better, than looking at (longer term)

monetary outcomes by themselves, or even better jointly. This latter is not

demonstrated by his model. Thus he attacks the Gerlach (2003, 2004) two pillar

Phillips curve estimate of inflation by arguing that the cointegration of inflation and

money means that money growth must be correlated in the longer run with inflation;

but the real question is whether attention to monetary trends adds anything beyond

what is also already visible in inflationary trends.

8

If two variables are cointegrated, then if they begin to deviate, there must be some

forces, or factors, restoring the relationship. But these forces, or factors, may impinge

primarily, or even perhaps solely, on one, or other, of the two variables. One reason

for the emphasis that monetarists place on the stylized fact that this long-run

relationship between monetary growth and price inflation continues to hold despite

there having been many differing money supply regimes is that it makes it much more

difficult to believe that the error correction mechanism is solely, or overwhelmingly,

from money holdings adjusting passively to pricing developments, rather than vice

versa. Of course, this is not a completely knock-down argument. Even under the

gold standard there were numerous forces leading monetary growth to adjust to

transaction requirements, for example via incentives to find more gold and/or

financial substitutes for gold when monetary conditions had become tight. Even so,

the likelihood that all such adjustment has been via passive monetary accommodation,

without any adjustment of inflation to monetary forces, especially in hyperinflationary

circumstances, seems highly improbable.

A subsidiary argument is that it is really inflation, and not the statistics on the rate of

growth of M3 or M4, that we really care about. So, even if there should be some

effect of excess money balances, (in the sense of money balances well above that

consistent with desired low inflation and sustainable output), on subsequent output

and inflation, we can still wait to see if it does actually feed through into higher

inflation, and then take countervailing action.

But even if one should agree with this ‘wait and see’ argument, the faster past growth

of money should at least be a warning that the future monetary policy decisions of the

9

Central Bank might need to be more restrictive, in the sense of higher interest rates

held for longer, than might otherwise be the case. Furthermore the ‘wait and see’

position depends on a number of arguable propositions, for example,

1. that demand shocks to money holdings are much more prevalent (and longer

lasting?) than supply shocks;

2. that lags between monetary policy action and its effect on inflation are short

enough that one can afford to wait until inflation actually appears in the data;

and/or that the extent of monetary action necessary to stabilize inflation (once

it has started – after a lag – to move away from target) will not then destabilize

the financial system and/or the real economy;

3. that we can, and do, measure inflation correctly.

Unlike most other economists, I do have doubts whether we measure inflation

correctly. A question arises how far, if at all, asset price increases, especially of

housing prices, ought to enter the index from which inflation is measured. For a

variety of reasons, monetary expansion may be more closely related to asset price

inflation, than to the inflation of goods and services prices. If inflation is (incorrectly)

measured, to exclude all such asset price inflation, then the links between money

growth and (true) inflation may be understated.

Moreover, in the longer run there should be cointegration between wealth and

income/expenditures. In so far as monetary fluctuations are closely associated with

those in wealth-holdings, then the resultant disturbance to the wealth/income ratio is

likely to have consequent effects on income flows and expenditures, and in due course

on goods and services inflation.

10

At this point I begin to move away from the simple three/four neo-Keynesian

equation system with which we began. The first amendment, naturally, from what I

have been saying already, is to put asset prices/wealth into the model. For obvious

reasons, the wealth/income ratio should be an argument in the expenditure function,

and wealth should probably be the scale variable in the demand for money function.

Finally there should be asset demand and supply functions, where the demand for

assets may in turn be a function of shocks to the supply of money, as well as to

expectations of future earnings, of future monetary policies and of future asset prices

themselves.

That said, I doubt if anyone connected with monetary policy would deny the effect of

the housing market, and of the equity market, on the forecast future for output and

inflation. But why should we go any further than that, and link the process back to

monetary growth specifically. This also raises the much debated issue of whether

Central Banks, or anyone else, should be concerned about asset prices, except as they

impinge on forecast values of real output and inflation.

Amongst the main arguments against using monetary policy to offset asset price

fluctuations are:-

I. that asset prices do not all move together in lock-step, and

II. that a member of an MPC is never in a strong position, ex ante, to claim that any

particular asset price is out of line with fundamentals (a bubble).

But so long as there is a reasonably close relationship between monetary growth and

asset prices in general, at least over some periodicities, these objections can be side11

stepped. The authorities are then responding to an excessively fast rate of growth of

monetary balances in general, and not to any particular set of asset prices, for example

to M3 growth in the eurozone rather than to housing prices in Spain, a point that

Otmar Issing has made several times (2002 and 2005).

Let me turn, finally, to my main point. This is that the so-called neo-Keynesian basic

model is based on inter-temporal utility maximisation by a representative agent, based

on the assumption that all debts are ultimately paid in full, otherwise known in the

jargon as the transversality condition. But this means that everyone is perfectly credit

worthy. Anybody’s IOU can, and would, be accepted in exchange. There is no need

for commercial banks, and there are none in Woodford’s iconic book, Interest &

Prices . Indeed it is hard to see why there should be any need for a specific monetary

asset, since everybody’s IOUs can be used for exchange purposes. All fixed-interest

financial assets are effectively identical, and there is one single interest rate in any

period, though it may shift over time as borrowing and savings propensities alter.

Moreover nobody, and no firm, is liquidity constrained, ever. Indeed the conditions

necessary for a no-default system to operate, either complete financial markets for

every possible contingency or perfect information, are, I believe, identical to those

that will allow a full Arrow-Debreu-Hahn Walrasian equilibrium to operate. As we

know, money is not necessary in such a system.

Thus, by basing their model on the transversality condition, the Neo-Keynesians are

turning their model into an essentially non-monetary model. So it is no surprise that

monetary variables are inessential in it. In reality, many agents in the economy, both

persons and smaller companies, cannot sell assets, since they do not have sufficient

12

saleable assets, or borrow, except at exorbitant interest rates. They are effectively

liquidity constrained, with their expenditures limited to their current income and their

few current assets. As Maurice Peston write in his 1980 book (Chapter 1),

“In Keynes’s General Theory consumption is determined by income to a very

considerable extent because the latter constrains the former. The poor

household has no liquid or marketable assets and can hardly borrow. It can

only spend its income.”

It is that constraint that modern Neo-Keynesian theory assumes away. Perhaps as we

all become richer, and come to own more assets, such constraints will in practice bind

less and less, and then money and commercial banks – and traditional Keynesian

analysis – will indeed become less important. But I do not believe that that time has

yet come. For a recent excellent empirical article on this, see Nier and Zicchino,

(2006). For the time being, the degree to which the current income, plus liquid asset,

constraint bears on current expenditures depends to a considerable extent on the

willingness of, and the terms on which, banks will lend to the private sector. This is a

key reason why I believe that the rate of growth of bank lending to the private sector

is as, or a more, important monetary aggregate than broad money by itself. Obviously

it makes no real difference whether an established company sells a bond to, or raises a

loan from, a bank, but a small company, or person, can usually only borrow from a

bank, and then only in loan form.

So, shifts in bank willingness to extend such loans, as banks become more, or less,

risk averse, will have the effect of shifting the constraints affecting the economy. In

13

particular, when the growth rate of the money stock is declining, whole segments of

the economy that were previously not income constrained may suddenly become so,

and at a time when income is probably also dropping.

Furthermore, when default becomes possible, risk premia come into play. There

ceases to be one single interest rate, as in the basic Neo-Keynesian model, but a whole

schedule of interest rates, depending on the perceived riskiness of the borrower.

Generally in depressions interest rates on safe, liquid government debt instruments

decline, but risk premia rise. It can then be ambiguous whether, overall, interest rates

have risen, or fallen. The reverse is true in booms; the official policy rate may rise,

but risk premia may fall. Against this background it would be short-sighted not to

cross-check for the combined effect that a combination of official policy measures

and changing risk aversion may have by looking carefully at the time path of the

monetary aggregates.

As Lord Peston asked (ibid), again in Chapter 1, “How far was the problem of

[achieving and maintaining] full employment one of dealing with or failing to cope

with risk and uncertainty.” I agree, but a measure of the willingness to face such risk

and uncertainty is given by evidence of the growth rates of the money and credit

aggregates. Keynesian economics emphasised income constraints and risk and

uncertainty. I have argued here that evidence on how the economy is coping with

these factors can be given by examining the growth rate of the money and credit

aggregates. In my view anyone who believes that default, risk aversion and income

constraints matter, whatever brand of Keynesian or Monetarist, ought to concern

themselves with the messages emanating from the monetary aggregates. To be sure

14

such messages are often garbled by noise, especially from short-run demand shocks,

so that such interpretation will be an art. Nevertheless it is an art worth attempting.

Let me now conclude with trying to set out a decision-tree on how to respond to

monetary data. By now the diagram should be easy enough to follow. The harder

part, no doubt, is to know at any time exactly in which box we find ourselves.

15

t – πt = b1dyt + b2dRt + dwt (5)μt – πt is predicted to be stationary.” So the growth rates

Is monetary growth consistent with

the current paths of y,

Yes

π and i?

No extra information

Consistent

with path of

Assets A

Do you want to take

action to alter A and

M?

No

Demand side

shock

No action needed

Supply side

shock

Bank behaviour

likely to affect y and

π

some extent.

16

Bibliography

Gali, J., (2006), Discussion of ‘Money and Monetary Policy: the ECB Experience

1999-2006’, by B. Fischer, M. Lenza, H. Pill and L. Reichlin, presented at the Fourth

ECB Central Banking Conference, ‘The Role of Money: Money and Monetary Policy

in the Twenty-First Century’, November 9/10.

Gerlach, S., (2003), ‘The ECB’s Two Pillars’, CEPR Discussion Paper, 3689.

Gerlach, S., (2004), ‘The Two Pillars of the European Central Bank’, Economic

Policy 40, 389-439.

Goodhart, C.A.E., (2005), `The Monetary Policy Committee’s Reaction Function: An

Exercise in Estimation’, Berkeley Electronic Press, Topics in Macroeconomics (5),

. Adjust policy to

http://www.bepress.com/cgi/viewcontent.cgi?article=1240&context=bejm

Issing, O., (2002), ‘Monetary Policy in a Changing Economic Environment’, in

Rethinking Monetary Policy, a Symposium organised by the Federal Reserve Bank of

Kansas City.

Issing, O., (2005), ‘Monetary Policy and Asset Prices’, in Schwerpunktthermen und

Serien, Börsen-Zeitung and Barclays Capital booklet, (Verlag Börsenzeitung).

McCallum, B., (2001), ‘Monetary Policy Analysis in Models without Money’,

Federal Reserve Bank of St Louis Review, 83 (4), July/August, 145-160.

Nier, E. and L. Zicchino, (2006), ‘Bank Weakness, Loan Supply and Monetary

Policy’, Bank of England, Work in Progress, (September 5).

Peston, M., (1980), Whatever happened to Macro-economics?, (Manchester

University Press).

The Times, (2006), ‘Obituary of Milton Friedman’, Times Newspaper, (November

17), p.

Uhlig, H., (2006), Discussion of ‘How Important is Money in the Conduct of

Monetary Policy?’, by M. Woodford, presented at the Fourth ECB Central Banking

Conference, ‘The Role of Money: Money and Monetary Policy in the Twenty-First

Century’, November 9/10.

Woodford, M., (2003), Interest & Prices, (Princeton, N.J., Princeton University

Press).

Woodford, M., (2006), ‘How Important is Money in the Conduct of Monetary

Policy?’, presented at the Fourth ECB Central Banking Conference, ‘The Role of

Money: Money and Monetary Policy in the Twenty-First Century’, November 9/10.