Money, credit and bank behaviour: need for a new approach.
The standard approach, in teaching and textbooks, to explaining the determination of both the supply of money, and the provision of bank credit to the private sector, has been the money multiplier approach, whereby the Central Bank sets the high-powered monetary base, and then the stock of money is a multiple of that. The greatest book on Monetary History ever written, Friedman and Schwartz (1963), Monetary History of the United States, was constructed around this same analytical framework of the money multiplier, whereby M, the money supply, would increase by a large multiple of the change in the high-powered monetary base, H.
M = H x (1 + C/D)/(R/D + C/D)
Yet when the authorities in the major developed countries attempted to use this relationship to expand the money stock (and bank lending) by force-feeding the banks with base money (H), in the process of Quantitative Easing (QE) in 2009, the prior relationships collapsed.
Prior to the commencement of quantitative easing in the UK and the US, starting in 2009Q1, and the aggressive expansion of liquidity by the ECB in the Euro Area, from 2008Q4 onwards, both the reserve/deposits ratio, R/D, and the cash/deposit ratio, C/D, had seemed to be fairly stable over the previous decade.
Moreover, the high levels of the C/D ratio, in the US and Euro Area particularly, were largely owing to foreign holdings and to black and grey market usage. Assuming these latter holdings to be invariant to domestic monetary expansion, then the true, underlying, money multiplier could be expected to be even higher. The higher level of the R/D ratio in the UK in June 2008, than the previous decadal average, was due to the fact that British commercial banks could choose their own R/D ratio, and by June 2008, nearly one year into the crisis, they chose to be much more liquid. In contrast, in Japan, the earlier years contained the aftermath of their own initial exercise in QE, from March 2001 to July 2006, whereas by June 2008 their R/D ratio had reverted to a more normal level.
Be that as it may, if we take either the average R/D, or the end June 2008 R/D ratio, and just feed the RID and C/D numbers into the money multiplier formula, the resulting values of the multiplier in these countries should have been somewhere around ten to fifteen, see table 2.
Instead, of course, the value of the multiplier declined so fast that the increase in both broad money and in bank lending to the private sector consistent with the huge increase in bank reserves was almost zero; in other words the incremental money multiplier approached zero, see table 3.
Quantitative easing occurred because the 'official' short-term interest rate had, approximately, hit the zero interest rate lower bound and so, in several cases for the first time ever, the authorities switched from using the official interest rate as their intermediate instrument to using the banks' reserve base as their target. The, largely unexpected, outcome, whereby the value of the multiplier then collapsed, is a nice example of Goodhart's Law, whereby, when the authorities try to make use of a prior statistical regularity as the basis for some new policy, then that regularity will collapse.
In the remainder of this paper, we shall first ask, in section 2, how such a major analytical failure was allowed to persist for so long. Then, and more importantly, we shall consider, in section 3, what kind of analysis should be introduced in its place.
2. The old approach
The so-called money multiplier is not derived from a behavioural relationship, but from a manipulation of identities:
M [equivalent to] D + C (1)
where M is the money stock, D the included set of bank deposits, and C is the cash holdings of the public.
H [equivalent to] R + C (2)
where H is the high-powered monetary base and R represents the reserve base of the set of included banks, their vault cash and deposits held with the central bank.
Divide all by D, and then (1) by (2)
M = H x (1 + C/D)/(R/D + C/D) (3)
So, equation 3 is an identity, not a behavioural theory, and has to hold at all times.
In fact, the actual behaviour of central banks was the reverse of what was usually (implicitly) assumed. This (implicit) assumption was that the central bank unilaterally fixed the amount of H, which then, via the money multiplier, determined the money supply, [M.sub.s]. So the level of short-term interest rates had to vary in the money market to equate the demand for money with the supply of money. This, market determined, short-term nominal rate then fed through into the other macroeconomic equations in the macro system.
This assumed monetary process was, of course, the exact opposite of how central banks have operated in practice. Quite why macroeconomists so commonly refuse to look at empirical observations in their theoretical formulation is remarkable, but beyond the scope of this paper (those interested might like to read Goodhart, (2009). Instead of setting H, the central bank sets its official interest rate, [i,sub.O]. Given [i.sub.O], the behaviour of the banks and of the private sector (to be discussed further in section 3) determines M, C, D and R. This determines the level of H consistent with [i.sub.O]. It is the job (normally) of the central bank money market desk to undertake Open Market Operations (OMO) so that [i.sub.M] = [i.sub.O], i.e. to set H consistent with M and [i.sub.O], where [i.sub.M] is the market short-term rate. Thus the so-called money 'multiplier' is actually a monetary 'divisor', determining the high-powered monetary base that the Central Bank has to provide, in order to make the short-term money market rate consistent with the chosen level of the official interest rate, i.e. making the official rate 'effective'; for a nice historical study of how the Bank of England did so, see Sayers (1957 and 1976).
The realisation that the money multiplier analysis was got comprehensively back to front raises a number of questions. We shall now consider four such questions:
i) Why did the authorities choose to operate in this way?
ii) How much did it matter that macroeconomists got the analysis back to front?
iii) Why did commercial banks choose to amass such massive amounts of cash reserves in 2009/10 rather than use such funds to buy assets?
iv) How does remunerating such reserves, rather than leaving them at a zero interest rate return, affect future (policy) developments?
Why did the authorities choose to operate in this fashion?
The central bank could, in theory, have operated to fix the total amount of H, or of its liabilities, rather than fix [i.sub.O], but except in a very few, and rather special, cases never did so. Recall that central banks were first founded during the age of metallic (gold) standards, and that their remit was, in part, to maintain that standard. The standard way to do so, to protect gold reserves, was to raise interest rates. Reducing assets/liabilities, e.g. by open market operations, was part of the process of raising interest rates, not an end in itself.
By the same token, the financial variable in expenditure functions, e.g. the IS curve, and in balance of payments, or capital flow, equations, is an interest rate, not a monetary quantitative aggregate. Observers have always tended to assess the effect of monetary policy on macroeconomic variables in terms of relative prices, interest rates, rather than of monetary aggregates. Moreover innovation, and changing financial structures, have frequently led to changing and uncertain definitions of what might at any time be the key monetary aggregates, whereas there has been much less temporal change and uncertainty about the definition of the risk-free (official) short-term interest rate.
Next, the demand for banks' cash reserves at the central bank was quite inelastic to the level of market rates, at least above a low level of the latter, so long as such reserves had zero remuneration, whereas
the demand for cash by the general public, and other factors, e.g. net payments to government and over the foreign exchanges, that affect the cash base of the system, could be very variable in a high frequency, e.g. day-to-day, context.
So, if the Central Bank kept R (bank reserves) constant and H varied, or if the demand for R and/or C varied (with a given H), then short-term market interest rates could, and would, become most volatile in that period (see figure 1, where the two vertical lines, [H.sub.1] and [H.sub.2], represent two (fixed) levels of H).
Since the official interest rate could be adjusted, and respond, to medium-term deviations of ultimate objectives (e.g. inflation, the exchange rate), from their target (such as a Taylor reaction function), there was no advantage, rather the reverse, from such high frequency short-term volatility. Indeed, it was hardly consistent with an overdraft, contingent claim, banking model, wherein a bank client could adjust his borrowing (up to an agreed limit) or deposits with a bank in an amount and at a timing of the client's choice, but at a prenegotiated rate of interest, relative usually to the official short-term rate. If this latter rate was, however, to be varying in the short run like a yo-yo for inexplicable reasons, this system could hardly have worked. Indeed, almost throughout their history, a crucial role for central banks was to be transparent about the factors that would cause them to vary their own rate. That remains so today. One purpose of such transparency is to facilitate the smooth operation of our banking and financial systems.
[FIGURE 1 OMITTED]
Did it matter that most economists got the analysis back to front?
The money multiplier analysis became adopted in leading academic circles in the 1930s (see Phillips, 1920; Keynes, 1930; and Robbins, 1932), and became incorporated in standard macro textbooks after World War II. Such a money multiplier approach has remained the centrepiece of analysis of money supply virtually to the current day. After the events of the past year, which surely demonstrated its unreliability, and given that it was a mistaken approach in the first place, I would hope that it would now be jettisoned.
This simplistic approach did not go without criticism. The seminal article, contrasting the standard money multiplier with the 'new' or behavioural adjustment approach, was Tobin's 1963 paper on 'Commercial banks as creators of "money"', though Tobin there (footnotes 2 and 3, p. 410) points to earlier work by Harry Johnson (1962), Gurley and Shaw (1960) and himself with Bill Brainard (1963) as having led the way. Maybe, but nowhere else was the argument set out so lucidly.
My own copy of the book in which this article appears, Carson (1963), has my annotated comment for Tobin's paper in the contents page, 'superb' (in contrast to a few other papers which I then rated as 'nuts', or 'bad', or 'unconvincing'). I was a convert. Since then I have continuously tried to promote the new view, in contrast to the standard multiplier approach, both in my earlier books (e.g. 1975, Chapter 6; 1984, Chapters 6 and 7) and throughout (2009, as well as here).
But rather like the mythical vampire, the money multiplier approach has never been completely and finally buried. Perhaps now?
Did it matter that the standard money multiplier analysis was back to front? After all, macroeconomics has survived for over 60 years in the face of such confusion (see also Disyatat, 2008). Of course one consequence was a persistent internal inconsistency between economists' policy advice and commentary on the one hand, and their formal analysis and pedagogic role on the other. With respect to the former, economists understood perfectly well that the choice was the level of interest rate to be set, while M and H were (largely) endogenous variables. But in their latter role they professed the opposite, that the central bank set H (and via the multiplier M), so that the riskless short-term interest rate became market determined. It was not until the Taylor reaction function became widely adopted (1) in the 1990s that this muddle was resolved.
It did have some malign effects on policy advice. For example, it led many to believe that the adjustment of reserve requirements, via calls for Special Deposits, etc., could have a significant impact on monetary growth. But, of course, after such a call for additional reserves to be held by banks, the central bank had to create them in order to keep market rates of interest in line with the official rate. This is not to claim that (calls for additional) reserve requirements had absolutely no effect on the money supply. So long as such requirements have zero, or at least below market, remuneration, they acted as a minor tax on banking, and hence in general did reduce the size of the banking sector slightly.
Of course, when the zero lower bound for interest rates was reached, there was then no limit to the scale of bank cash reserves that could be created. So, now for once, the money multiplier could be used in the fashion that academic economists had assumed, with the central bank injecting H and R by large-scale open market operations. Initially at least, some thought that the multiplier might then work, as supposedly designed, to generate a large-scale resultant increase in M and in bank assets. It did not work that way; instead banks were content to amass, by historical standards, huge amounts of reserves at the central bank over and above any residual regulatory requirements.
Why did banks not use their large reserve holdings to purchase assets in 2009/10?
The central bank is the monopoly supplier of cash and bank reserves. Subject to a few minor qualifications, which do not matter for our purposes here, when the central bank has injected H and R into the banking system, the commercial banks in aggregate cannot avoid holding that total amount of reserves. So the overall volume of reserves is determined by the central bank, not by the commercial banks (Keister and McAndrews, 2009).
While this is indeed so, each individual commercial bank, should it feel that it held more cash reserves than it needed, could aim to rebalance its own portfolios by buying some asset in the financial market, thereby setting the supposed money multiplier in operation. The remarkable feature of the recent episode of quantitative/ credit easing was that this money multiplier totally failed to work, as set out in the textbooks. Instead of responding to the massive injections of additional reserves by going out and buying new assets or making new loans, the commercial banks seemed happy just to sit on their massively increased cash base, with virtually no credit expansion. Why?
Commercial banks seek to maximise risk-adjusted profits, subject to various regulatory constraints, such as capital requirements. In 2009/10, in the middle of a severe recession, the risks of lending to the private sector appeared high and rising; banks' own risk aversion had become enhanced and regulatory requirements, on leverage, capital and liquidity, were in the process of being raised. Under these circumstances most banks severely tightened the terms on which they would provide access to borrowing facilities; the extent of collateral cover, the spread of loan rate over the official rate, the loan to value ratio, etc., were all made tougher. With housing, and other asset prices, often declining, and little, or no, investment demand in the downturn, there was little demand to take on additional bank loans on these tougher terms. Indeed large corporations took the opportunity to borrow in capital markets, and to repay bank borrowing, so as to strengthen their balance sheets.
But if commercial banks found little, or no, demand for loans from the private sector, at least on the terms that they are now demanding, why did they not expand their purchases of public sector debt, especially since this generally had a zero risk weighting and would count as a liquid asset, with required liquidity ratios being prospectively on the increase?
But the context was one in which official short rates had been brought down to zero, while public sector deficits and debt/income ratios were rocketing upwards in an unprecedented fashion. There was, indeed, increasing concern about sovereign risk, with some facets of that concern being more justified than others. The clear implication was that expected future interest rates would be higher than current spot rates. Meanwhile central banks, in the US and UK, began to offer remuneration on all deposits held with themselves, not just for required reserves, at a rate equal to, or just slightly below, the official rate. Thus the choice for banks was between a remunerated riskless deposit at the central bank and holding public sector debt with a slightly higher return but subject to potentially severe adverse interest rate risk. The authorities had created a liquidity trap for themselves! They did not appear to understand what they had done in this respect, or to appreciate how policy might respond to such circumstances.
Now that bank reserve deposits with the Central Bank are remunerated, how might that influence policy?
Almost all the main central banks of developed countries have now moved over to a 'corridor' system, with the Federal Reserve System being the last to adjust. Under such a corridor system, the official rate sits between (and usually exactly in the middle of) an upper band, at which rate banks can obtain unlimited extra cash from the central bank at their own volition (so long as they can provide the requisite collateral), and a lower band at which they will be remunerated for (unlimited) deposits with the central bank.
Under normal circumstances the width of the band is traditionally set at a round number, such as plus or minus 1 per cent around the official rate. The central bank money market desk then aims, via open market operations, to keep short-term market rates close to the official rate, and the bands only operate to dampen short-term market interest rate volatility caused, for example, by unforeseen cash flows. In other words, policy proceeds much as usual, with or without bands.
But the existence of the corridor, and its bands, gives the central bank an additional instrument. If the corridor is widened, with, say, the spread to each band doubled, application to the central bank for resolution of dislocations in financing flows will be discouraged. So the bandwidth will tend to be widened (narrowed) when wholesale, e.g. interbank, private sector markets are working well (badly), and when the central bank has few (many serious) concerns about the provision of liquidity to the banking sector. Not surprisingly, corridor bands were generally narrowed during the recent financial crisis.
Moreover, if banks have persistently excess (insufficient) reserves, relative to requirements, then the market rate will gravitate to the lower (upper) band of the corridor, so long as open market operations are not used to force market rates back into line with a positive official rate. It may be that the authorities want to adjust interest rates and bank liquidity separately; an example might be a desire to raise rates to stem a collapse in the foreign exchange market at the same time as being unwilling to impose extra debt sales on a fragile government debt market. With a corridor system, this can now be done. The authorities simply raise the whole level of the corridor (see figure 2), with official rates raised from [i.sub.O1] to [i.sub.O2]. Given central bank's ultimate control over the cash base, interest rates can in practice be moved by fiat, (open mouth policy), with little, or no, need for accompanying open market quantitative operations.
If there were excess cash reserves at level 1, so that market rates would be bumping along the bottom corridor, [L.sub.1], now after rates are raised to level 2, market rates would still be bumping along the lower band at [L.sub.2].
Most central bankers appear to regard it as necessary to keep the bands symmetrical around the official rate. Why they seem to think this is not clear, and it throws away an additional valuable instrument (see also Perez-Quiros and Mendizabal, 2010). Thus a more expansionary liquidity policy would involve asymmetrically lowering both the upper band (making it nearer to the official rate) and the lower band (making it further from the official rate), if necessary charging a fee--a negative rate of return--for keeping deposits at the central bank (Palley, 2010). Per contra, a restrictive liquidity policy would have the central bank raise both bands relative to the official rate.
[FIGURE 2 OMITTED]
[FIGURE 3 OMITTED]
No central bank has yet, to my knowledge, applied such an asymmetric policy, but the Fed is considering raising the deposit rate up to the level of the official rate, should it want to restrict bank expansion at a time when it does not want to undertake open market operations to reduce excess reserves. But as long as the marginal return, adjusted for risk, on bank asset expansion was perceived, by the banks, as higher than the deposit rate, then this instrument would be of relatively minor efficacy.
But the main purpose of this paper is not so much with policy prescriptions, but with the need for a new analytical approach to the determination of the supply of money, and it is to this that I now turn.
3. A new approach to the analysis of the determination of the money supply
A major drawback of the money multiplier approach was that it was almost entirely mechanistic with no behavioural context; both the commercial banks and the private sector acted as passive automata, not maximising anything. The only behaviour (implicitly) assumed related to the central bank, and that was back to front, assuming that their choice variable was H, whereas in fact it was [i.sub.O].
Having now reconsidered the central bank's choice variable, the first step along the road to an improved analysis is to review what are the choice variables (and what are the residual endogenous variables) of the commercial banks. Here I am simply going to assert, on the basis of observation, that what the banks set are interest rates, offered on deposits of varying durations, and charged on loans of different formats, e.g. overdrafts, term loans, etc., plus certain associated terms and conditions, such as penalties for early withdrawal or exceeding limits, collateral and down-payments required on mortgages, duration, etc. Given such rates and spreads, the amount of such deposits/ loans is then determined by the choice of the private sector, and cannot be controlled by the banks themselves.
Such commercial bank interest rates, on loans and deposits, are normally set as a spread relative to the official rate, so changes in the official short-term rate lead, subject to some short lags, to roughly equivalent changes on the pattern of rates applied by all (bank) financial intermediaries, so that the spread on deposits, SD, is the margin between the interest rate offered on deposits and the official short-term rate:
SD = [i.sub.O] - [i.sub.D] (4)
and the spread on loans, SL, is the margin between the loan rate and the official rate:
SL = [i.sub.L] - [i.sub.O] (5)
Such bank spreads can themselves be affected by longer-term structural factors, such as costs and productivity in banking, innovations in banking and in finance more generally, and the extent of competition both within the banking sector and between banks and alternative nonbank sources of finance. Besides such longer-term factors, that help to determine the 'normal' level of spreads, there are conjunctural, cyclical variables that help to determine whether banks are in an expansive, or contractionary, mode. Finally, given the existing spreads, the balance between loan and deposit expansion is determined by the private sector, not by the banks themselves. So, as the L/D ratio varies, one means of adjustment by the banks is to shift the relative spread on loans and deposits.
Amongst the set of conjunctural, cyclical factors that will help to determine spreads are the perceived, or expected, returns on new loans, and their expected riskiness, ER, E[sigma] [sup.2]R, together with such constraints as the regulators and prudence may impose on bank expansion in the form of capital and liquidity requirements. Clearly the higher expected profitability, the less perceived risk, and the less binding are (regulatory) requirements, the smaller such spreads will become, in order to expand leverage, hence:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (6)
(remembering that a negative partial implies a narrowing of the deposit spread) and;
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (7)
(where a negative partial again implies a narrowing of the loan spread).
In addition, banks have to respond to differential movements in loan demand and in deposit provision from the private sector; thus L - D will vary over time in ways that banks may not be able to foresee. As, for example, loan demand rises relative to deposit supply, then a natural response by banks is to raise the loan spread, so as partially to choke off such demand,
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (8)
and to lower the deposit spread to compete more vigorously for funding to finance such loans, hence
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (9)
Hence putting equations (7) and (8), (6) and (9) together, we get,
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (10)
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (11)
Up till this point, I have ignored a complication, which is that banks generally set the terms, rates and limits for access to credit, i.e. the provision of credit facilities, whereas both the timing and amount of usage of such facilities is at the command of the bank borrower (client). What the bank(s) can, more or less, control by varying terms, etc., is the approval of the new loan facilities; it cannot, to the same extent, control their usage, so SL influences credit lines extended, not usage.
Available, accurate data are, however, for ex post usage, not for ex ante facilities; the definition of these latter, moreover, is fuzzy, and the perceived probability of usage of various kinds of lines of credit can vary greatly. Moreover, credit extension and credit usage can move in different directions. For example, after the post-Lehman panic in 2008Q4, new bank credit extension probably dropped sharply (a surmise since proper data are not available) whereas credit usage went up sharply, but temporarily, in many countries, e.g. the US, as borrowers with unused facilities sought to lock these in before banks could unilaterally withdraw, or reduce them (as my UK bank did to me, simply announcing that my personal overdraft limit had been cut from 10k [pounds sterling] to 3k [pounds sterling]). So the quantitative data necessary to test hypotheses about bank behaviour are, on this front, largely lacking. As a result, during this recent crisis much more weight has had to be placed on survey data, both from lenders such as Loan Officers and from potential borrowers.
Because variations in rates, terms and conditions affect facilities, rather than usage (for both loans and deposits), and because there are quite severe competitive constraints on any individual bank's ability to vary spreads on its own, variations in such spreads will never, by themselves, be able to bring loan extension and the growth of retail deposits into balance. Even when the banking system as a whole is roughly in balance, some banks will be net lenders (D > L) and others net borrowers (D < L).
So what are the adjustment mechanisms, operated by the banks' Treasury Offices, that enable banks' balance sheets to balance, when D [not equal to] L? There are two such mechanisms, one using the asset side of a bank's portfolio, the second using liability management. The first mechanism, via adjustment of liquid assets, was that adopted historically. Besides loans, when the bank sets rates, terms and conditions, and then accepts usage at the behest of the client with the facilities, banks can also purchase existing financial assets, usually some form of public sector debt, where (unlike the loan market) the banks are price-takers and quantity setters. So, when L > D, the banks can adjust by selling back such assets onto the market. The more these assets are liquid, the less risk the bank faces in adjusting to fluctuations in L and D. The determinants of bank liquid asset holdings include current liquidity requirements, trends in L - D, relative returns on holding such liquid assets, e.g. relative to loans, and the availability of other mechanisms of balance sheet adjustment (Other) so:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (11)
where [Q.sub.ps] is the quantity of public sector debt held, and [i.sub.ps] - y are the relative returns on holding such debt. In recent decades Liquidity Requirements have been progressively dismantled, L - D has shot upwards, [i.sub.ps] - y has declined somewhat, and alternative, liability management adjustment mechanisms burgeoned. Hence [Q.sub.ps]/D collapsed to tiny values.
The second main adjustment mechanism is, of course, liability management. Besides trends in L - D, this has been a function of the relative interest rate spread between wholesale rates, as represented by LIBOR, and the official rate, and by innovations in wholesale markets and techniques, so that:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (13)
Such innovations have been manifold in recent years, so much so that it was thought that, so long as banks kept a sufficient capital buffer to ensure solvency, they could always access such sufficient wholesale markets for almost unlimited sums. Meanwhile, until 2007 (L - D) trended strongly upwards and (Libor - [i.sub.O]) went down as risk spreads declined to low levels. Accordingly both the volume and share in total liabilities of wholesale deposits rose sharply, a key feature making the banking system more susceptible to the crisis of 2007-10.
Meanwhile, once banks have set the spread of deposit rates, relative to the official rate, SD = ([i.sub.o] - [i.sub.d]), then retail deposits are almost entirely demand determined in the standard fashion, so
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (14)
where y is real output, p the price level, [pi] the inflation rate, and [i.sub.n] is a set of rates on other alternative assets; the partial derivatives take standard signs.
In this equation, SD, the deposit spread, is a function of supply-side factors, equation (7), but SD is normally fairly stable, determined by longer-term structural factors. Otherwise retail deposits are primarily demand-determined, though perceptions of liquidity (and bank) risk and wealth may also enter. Insofar as retail deposits are purely demand determined, then by the same token this equation becomes superfluous, in a neo-Keynesian three equation model, as in Woodford (2008):
[y.sub.t] = [Ey.sub.t+1] + [[beta].sub.l]([i.sub.t] - [E[pi].sub.t+1]) + [u.sub.t] IS (15)
[[pi].sub.t] = [E[pi].sub.t+1] + [beta].sub.2][y.sub.t] + [v.sub.t] LM (16)
[i.sub.t] = [Florin]([E[pi].sub.t+1] + [Ey.sub.t+1]) Taylor Reaction Function (17)
In contrast to retail deposits, wholesale deposits are clearly driven as much, or more, by supply factors as by demand factors. There are often huge fluctuations in the broad money holdings of the shadow banking system, of money market mutual funds and of non-bank financial intermediaries; these are often driven not only by the need to fund lending growth in excess of the supply of retail deposits, but also by obscure (tax and regulatory) arbitrage considerations that influence where deals and transactions are booked. This means that the interpretation of fluctuations in wholesale broad money data can often be difficult.
So after this tour d'horizon of money-supply determination, how do we end up assessing the meaning and implications of the various monetary aggregates:
1) H, R and C. Not worthy of much consideration. H and R are endogenous under normal conditions, and have an unstable relationship with M and L at the zero interest bound. C is also almost entirely demand determined, but may at times carry information about concerns about bank safety, or the growth of the black economy, and can sometimes be a coincident indicator of consumption expenditure.
2) M1, narrow money. Demand determined. Can be very interest elastic, especially when rates on competitive time deposits fall very low. So M1 data tend to exaggerate how expansionary monetary policy has become in slumps, when all interest rates tend towards zero, and how restrictive monetary policy is in a boom, when relative interest rates rise. So M1 data provide excessive comfort to the monetary authorities, and hence are something of a snare and delusion for them.
3) Retail Money, i.e. cash plus the bank deposits of the household and non-financial corporate sector. This may be the nearest available statistic to the older concept of broad money, but, in order to obtain such data, it is necessary to have a sectoral split of bank deposit holdings, which is not (or no longer) available in many countries. This is primarily demand determined, but it certainly can be influenced both by relative interest rates, being a negative function of the spread between loan and deposit rates and, perhaps, a negative function of the ease of access to additional credit facilities. So fluctuations in this sectoral aggregate can be influenced by supply-side factors, and need careful interpretation, but probably give the most informative quantitative indication of the general monetary conditions of the non-financial private sector.
4) Broad Money. This has been massively distorted in recent years by supply-side conditions, notably liability management to bridge the gap between L and D. But beyond that there have been large, often huge, shifts of funds between the banking and shadow-banking sectors, often driven by tax and regulatory arbitrage. Whereas some of the fluctuations in the monetary holdings of non-bank financial intermediaries may be economically important, e.g. as part of the transmission mechanism from liquidity injection and lower short-term interest rates towards higher asset prices and lower yields on riskier and longer-term assets, the interpretation of such data remains quite complicated, even arcane. In the US the Fed has given up publishing such data, as too impenetrable, and the Bank of England places more weight on the sectoral figures.
5) Bank credit to the private sector. As noted earlier, the availability of access to cash is probably more important (to borrowers) than the usage of such facilities. Moreover credit can be provided by other financial intermediaries and by suppliers, in the guise of trade credit, as well as by banks. The banking system can provide money to the private sector by lending to the public sector, and even to overseas residents, as well as to the private sector. So, while bank lending to the private sector, and the ease of access to additional such loans, are important data, they too need interpretation and assessment.
To sum up this discussion of the various monetary aggregates, none of them in isolation can be regarded as a sufficient, or entirely satisfactory, statistic of monetary conditions in the large. All such monetary data need assessment and interpretation before forming a judgement about monetary conditions and their likely effect on the real economy. But the difficulty of doing so should not stop the effort being made, because interpretation of monetary quantitative data is an important adjunct and complement to the focus on interest rates.
The old pedagogical analytical approach that centred around the money multiplier was misleading, atheoretical and has recently been shown to be without predictive value. It should be discarded immediately.
The practical realities, whereby the central bank and the commercial bank set the interest rates at which they will operate, and then the various agents in the private sector and amongst the banks determine monetary quantities endogenously, is more complex but has the advantage of realism. Not only are such monetary mechanisms quite complex, but they are changing quite rapidly, especially in response to the pressures of the recent financial crisis. Moreover the regulatory response to that crisis is likely to toughen financial regulations, making the banking sector smaller as well as safer. Just how the banking sector, and the real economy, will respond to such pressures remains to be seen.
Much more analytical work on this subject needs to be done. Ars longa, vita brevis.
Brainard, W. and Tobin, J. (1963), 'Financial intermediaries and the effectiveness of monetary controls', American Economic Review, 53, May, pp. 384-86.
Carson, D. (ed.) (1963), Banking and Monetary Studies, Homewood, Ill., Richard D. Irwin Inc.
Disyatat, P. (2008), 'Monetary policy implementation: misconceptions and their consequences', Bank of International Settlements Working Paper 269, December.
Friedman, M. and Schwartz, A.J. (1963), A Monetary History of the United States, 1867-1960, National Bureau of Economic Research, Princeton, Princeton University Press.
Goodhart, C.A.E. (1975), Money, Information and Uncertainty, London, Macmillan Press Ltd.
--(1984), Money, Theory and Practice: The UK Experience, London, Macmillan Press Ltd.
--(2009), 'The continuing muddles of monetary theory: a steadfast refusal to face facts', Economica, 76, pp. 821-30.
Goodhart, C., Baker, M. and Sleeman, C. (2009), 'Quantitative easing: necessary, successful and ready to wind down', UK Economics, Morgan Stanley Research Europe, 23 November.
Gurley, J. and Shaw, E.S. (1960), Money in a Theory of Finance, Washington, D.C., The Brookings Institute.
Johnson, H. (1962), 'Monetary theory and policy', American Economic Review, 52, June, pp. 335-84.
Keister, T. and McAndrews, J. (2009), 'Why are banks holding so many excess reserves?', Federal Reserve Bank of New York Staff Report 380, July.
Keynes, J.M. (1930), A Treatise on Money, London, Macmillan and (1971) New York, St. Martin's Press.
Palley, T. (2010), 'The troubling economics and politics of paying interest on bank reserves: a critique of the Federal Reserve's exit strategy', paper prepared for a session titled 'Countercyclical Macroeconomic Policies' held at the meetings of the Eastern Economic Association held in Philadelphia, PA, February 26-28.
Perez-Quiros, G. and Mendizabal, H.R. (2010), 'Asymmetric standing facilities: an unexploited monetary policy tool', Banco de Espana Documentos de Trabajo 1004.
Phillips, C.A. (1920), Bank Credit, New York, Macmillan.
Robbins, L. (1932), An Essay on the Nature and Significance of Economic Science, London, Macmillan.
Sayers, R.S. (1957), Central Banking After Bagehot, Oxford, Clarendon Press.
Sayers, R.S. (1976), The Bank of England 1891-1944, Cambridge University Press.
Tobin, J. (1963), 'Commercial banks as creators of "money"', Chapter 22 in Carson (1963) op.cit., pp. 408-19.
Woodford, M. (2008), 'How important is money in the conduct of monetary policy?,Journal of Money, Credit and Banking, 40, 8, pp. 1561-98.
(1) John Taylor was not the first economist to consider and assess empirically to which variables central bankers responded when they set interest rates, i.e. the form of their reaction function. It would be an interesting exercise in the history of economic thought to explore the earlier history of studies on central bank reaction functions and, in particular, what made such earlier work go relatively unnoticed, whereas Taylor's formulation became standard throughout the profession quite quickly.
C.A.E. Goodhart, Financial Markets Group, London School of Economics. e-mail: email@example.com. I am grateful to Rafael Repullo for advice and comments and to Nelson Camanho Costa-Neto for research assistance. All remaining errors are my own.
Table 1. R/D and C/D have looked fairly constant for a decade. R/D (as a percentage) Average Standard Level in deviation June 2008 UK (June 1998 to June 2008) 0.59 0.71 2.46 US (June 1998 to June 2008) 0.84 0.15 0.65 Euro Area (January 1999 to June 2008) 2.44 0.20 2.71 Japan (July 1997 to April 2008 1.45 1.07 0.68 C/D (as a percentage) Average Standard deviation UK (June 1998 to June 2008) 4.80 0.26 US (June 1998 to June 2008) 12.02 0.30 Euro Area (January 1999 to June 2008) 7.19 0.91 Japan (July 1997 to April 2008 6.41 0.80 Table 2. ... So, average multipliers should have been ... Using an Using a June average R/D 2008 R/D UK 19.45 14.44 US 8.71 8.84 Euro Area 11.14 10.83 Japan 13.54 15.01 Table 3. But tiny multipliers (PNFCs and the household sector), per cent June 2008- Change in Change in Change in June 2009 bank reserves broad money bank lending held at to private central bank sector UK 371 2 1 US 1853 9 4 Eurozone 122 4 2 Japan (a) 103 8 -17 Source: Goodhart et al. (2009). Note: (a) March 2001 to March 2006.
|Printer friendly Cite/link Email Feedback|
|Publication:||National Institute Economic Review|
|Date:||Oct 1, 2010|
|Previous Article:||Modelling the UK banking sector.|
|Next Article:||Unemployment: introduction.|