RBA Annual Conference – 1989 Overview: Monetary Policy and the Economy Ian Macfarlane and Glenn Stevens
It is easiest to discuss the performance of monetary policy when there is a generally agreed proximate objective for policy, which can then serve as a yardstick for judging its success or failure. A decade ago, anyone who sought to evaluate monetary policy had this luxury. There was not much debate about what the proximate objective should be, or about how to evaluate monetary policy. The influential framework of the revived quantity theory of money, together with high inflation, focussed attention on monetary aggregates as the intermediate objective, and the most important measuring standard, for monetary policy. The late 1970s and early 1980s represented what Goodhart[1] has referred to as “the high-water mark of national monetarism”. This followed, of course, a much longer period – the Bretton-Woods era of fixed exchange rates, which, in the limit, subordinated domestic monetary policy to exchange rate objectives.
As the 1980s proceeded, however, the aggregates-based approach to monetary policy came under pressure, failing to gain new adherents and losing some former practitioners. European countries strengthened their commitments to stable exchange rates in the European Monetary System. Other countries grappled with the increasing instability of traditional money/income relationships in the face of financial deregulation and international integration. Some countries stopped targeting altogether. Others either switched the aggregates they used for targeting purposes, or subtly downgraded their commitments to targeting.
In Australia, the policy of targeting M3 was abandoned in January 1985, largely because of the likely distortion of M3 caused by the commencement of new banks, and the continuing effects of earlier deregulation of existing banks (the last major step, to that date, being their entry into the overnight money market in 1984). Banks fought aggressively for, and won, market share from non-bank financial intermediaries. In the event, the growth of M3 in 1984/85 was almost twice the rate that had been originally projected. The judgment was made that the tightening of monetary policy to the degree required to have brought M3 within the projection would have been unjustified.
To some extent, this decision was made under the pressure of emerging events, without the benefit of exhaustive empirical research. Many policy decisions are. Subsequent research work in the Bank has re-examined, on a piecemeal basis, the usefulness of various financial aggregates.[2] The purpose of the present set of papers is to systematise this work, taking advantage of international experience and the further data becoming available on Australia.
The papers should be seen within the traditional framework for thinking about monetary policy, in which policy instruments are used in pursuit of an intermediate objective which bears a relationship to some ultimate objective. They address several questions about financial variables as possible intermediate objectives. Are demand functions for money less stable than formerly? Can more stability be gained by looking at broader aggregates? Do financial aggregates lead or lag economic activity? How have financial aggregates been affected by the changed competitive behaviour of banks? What is the role of interest rates in all of this? If aggregates cannot be used as intermediate objectives, how should monetary policy be conducted?
The plan of the volume is firstly to review the experience of countries that used monetary aggregates as intermediate targets. Then there are two papers which look at the relationships between various financial indicators and economic activity in Australia. These are followed by a paper which attempts to go behind some of the changes in banking aggregates (such as M3) and look at changes in banks' competitive behaviour. The fifth paper looks at the feasibility of conducting monetary policy with a view to tying down some nominal final objective, such as prices or nominal GDP, without the benefit of a monetary aggregate as an intermediate objective. Finally, there is a summary of discussion and comments by participants at a Conference held in June 1989 to discuss the papers.
This introductory essay is designed to explore the related themes running through the papers and, in a couple of places, to say more about the “model” which underlies the thinking in some of the papers.
The paper by Argy, Brennan and Stevens suggests that the adoption of monetary targets by the industrial countries in the mid 1970s was a tactical response to a particular situation: recent experience of high inflation, unstable inflationary expectations, and concern about the monetary implications of fiscal policy. In pursuing their stated monetary targets, most countries had regard to the state of the economy and other objectives, especially exchange-rate objectives. Monetary targets were relegated down the order of priorities when circumstances were deemed to require that. Most countries tolerated overruns of targets, some consistently, and there were few occasions where this was offset by lower monetary growth in subsequent years. In short, while the rationale for monetary targets owed a lot to the idea that a monetary rule was a good guide for monetary policy, in practice, monetary targets were never interpreted, nor were they intended, as such rules.
That is not to say that targets were of no value. But it is suggested by Argy et al that monetary policy should not be assessed on the basis of whether or not announced monetary objectives were achieved, but on the basis of the performance of the final objectives of monetary policy. That is much more difficult, since developments in inflation, output, etc. may reflect many factors other than monetary policy's influence, but it is a more fruitful line of enquiry. Much of the progress in reducing inflation was made in the 1980s, well after the introduction of targeting itself. Factors other than monetary policy were important: the decline in oil prices, and the absence of adverse shocks, for example. But tight monetary policies (especially in the United States) made their contribution. In defence of targets, it could be said that they were helpful, not because they had the characteristics of a monetary rule, but because they enabled monetary authorities to adopt tighter policies than might otherwise have been possible.
One common feature of the experience of targeting countries was the need to change the definition of money targeted, or to change the relative importance between multiple targets. The only countries surveyed by Argy et al which did not do this were those which dropped targets altogether. The reason for this (and the reason for dropping targets) was, in almost all cases, a deterioration in the relationship between money, as the intermediate objective of monetary policy, and nominal GDP or prices, as the (assumed) final objective of policy.
The paper by Bullock, Morris and Stevens sets out some stylised facts from the past two decades of Australian experience. Previous work had given an econometric treatment of the issue of the stability of the demand for particular monetary aggregates (Stevens, Thorp and Anderson (1987), Blundell-Wignall and Thorp (1987)). The strongest conclusion of this work was that the demand function for M3 had broken down in the 1980s. The Bullock et al paper conducts an examination of the relationships between a range of financial indicators – all the currently recognised monetary and credit aggregates, and short-term interest rates – and private final expenditure.
On the monetary aggregates, Bullock et al note that in terms of turning points, M1 leads private demand, regardless of whether the latter is measured in nominal or real terms. M3 had that characteristic in the 1970s, but has since tended to be at best a coincident indicator, and in one important episode (the 1982/83 recession) was a very poor indicator indeed. Broad money has tended to be quite well related to spending (Blundell-Wignall and Thorp (1987) had already concluded that the demand function for broad money was relatively stable), but in terms of timing, has tended to be coincident or lagging.
Bank lending is not a very useful indicator most of the time, and certainly is inferior to M3. That bank lending and M3 (which consists chiefly of bank deposits) have often had quite different short-term trends, reflected large changes in banks' holdings of liquid assets in the 1970s, and use of non-deposit liabilities in the 1980s. These issues are covered more fully in the paper by Battellino and McMillan.
Movements in broader lending and credit aggregates have a reasonably good relationship with the business cycle, with a noticeable lag. Although on some measures this relationship is more stable than for narrower aggregates, there is also some evidence that it has shifted in the 1980s, in the sense that a given rate of growth of nominal spending or output has been associated with a higher rate of growth of lending and credit than was formerly the case.
Bullock et al also note that there is apparently a relationship between short-term interest rates and growth in private spending, in the sense that periods where interest rates have risen sharply to very high levels have also been periods where private spending, especially when measured in real terms, weakened noticeably. The simple empirical evidence, if taken at face value, suggests that this relationship has strengthened in the 1980s.
The significance of the latter results lies in the fact that short-term interest rates are, for all intents and purposes, the first line in the process by which monetary policy is “transmitted” to the economy. This is the case not only in Australia, since deregulation of interest rates, but also in most developed economies. Given that many of the potential intermediate objectives for monetary policy – such as money and credit aggregates – have tended to give signals that are difficult to interpret, it is of more than passing interest if a relationship between what is effectively the instrument of monetary policy and its final objectives can be established.
Of course, it is real interest rates which, in principle at least, are important in the determination of expenditure and output. Most of the analysis in Bullock et al uses nominal interest rates, wary of the uncertainties in measuring real interest rates, and noting that for much of the 1980s, movements in real and nominal rates have probably been quite similar. Follow-up work, presented in this volume, suggests that if a simple measure of real interest rates is used in the same way that Bullock et al used nominal rates, the results are not too different. The same work suggests that a simple proxy for the slope of the the yield curve (the difference between 10-year and 90-day rates) is perhaps a slightly better indicator than the real short-term rate on its own.
The paper by Stevens and Thorp employs a different technique (“Granger-causality tests”) to that used in Bullock et al, on the same set of variables. The results there generally confirm that the broad monetary and credit aggregates lag, rather than lead, economic activity. Not much can be said, on the basis of these tests, about the narrow aggregate M1, or the intermediate aggregates M3 and bank lending.
The techniques in Stevens and Thorp do not shed much light on the relationship between interest rates and activity. There is no support for a leading role for interest rates, and there is some suggestion that there is a feedback from expenditure to interest rates, though this is quite sensitive to model specification.
These tests are quite demanding: previous information about one variable must explain the current value of another variable, after past values of the latter variable itself have been used. Any hypothesis which is not rejected in this sort of test can be held very firmly indeed – which suggests that we may have a fair degree of confidence in the idea that broader aggregates tend to lag. On the other hand, other hypotheses may well be rejected too readily on the basis of the Granger-causality tests. If we take the results literally, they would suggest that there is no leading relationship between any financial variable and real or nominal activity; activity can be explained entirely by its own past values.
A particular problem may occur when this methodology is used to test for the relationship between short-term interest rates and activity. This is because there is a two-way process at work. As activity speeds up, it is likely that monetary policy will be tightened and this will show up as a rise in short-term interest rates. In time, when the higher interest rates have had their effect, activity could be expected to slow down. There is thus a link from activity to interest rates and another from interest rates to activity. In these circumstances, it is unlikely that a result of one variable unambiguously leading another will be found.
The really important question, of course, is how the monetary “transmission mechanism” actually works. The answer to this question that would be given now is different to that which might have been given a decade ago. A change in monetary policy under the old regulated system would frequently have directly affected the level of M3. For example, a rise in the administered yield on government securities would induce the non-bank private sector to switch from bank deposits to government securities. Banks' deposits, and their liquid assets, would be immediately reduced. Without the freedom to compete actively for deposits, banks would thus be forced to curtail lending so as to maintain the required ratio of liquid assets to deposits. To allow some short-term flexibility, banks typically held large quantities of “LGS” assets in excess of the minimum requirements.
One complicating factor was the prevalence of overdraft lending, where the actual amount of credit outstanding was, to a considerable extent, determined by the borrower. As a result, bank loans outstanding could often be slower to respond to interest rate changes than was M3. But broadly speaking, rationing of bank credit was an important feature of the way monetary policy had its impact on the economy in the 1970s. This mechanism was also, of course, important in the housing sector, traditionally very sensitive to changes in financial conditions, until as late as 1986.
Much of this has now changed, as is pointed out in the paper by Battellino and McMillan. The deregulation of banks' interest rates in the 1980s has allowed banks to manage the liabilities side of their balance sheets much more actively. Banks are no longer in the position of passively accepting whatever inflow (or outflow) of deposits comes their way. They can determine the asset composition of their balance sheets more or less as they choose (subject still to the the Prime Assets Ratio), and then arrange the liabilities side of their balance sheets so as to fund the assets in the most cost-effective way. For much of the 1980s, this has involved recourse to foreign currency borrowings, and Australian dollar liabilities offshore (both of which avoided the implicit tax of the Statutory Reserve Deposit arrangement) as well as the use of bank bills, which do not require the bank to fund the loan at all (also avoiding the SRD).
In this world, it tends to be the assets side of the balance sheet which drives the liabilities side, and the composition of liabilities is prone to large, and very rapid, shifts according to relative funding costs. For all intents and purposes, the quantity of “money”, defined as M1, M3 or some other “M”, will be determined endogenously: there is no thought of the central bank actually directly controlling the supply of this “M”, as is assumed in the conventional textbook treatment, which describes the first stage of a change in monetary policy as “ΔM”[3].
The more accurate way to consider the starting point of the monetary transmission process is as a change in very short-term – i.e. overnight – rates of interest. This will usually be reflected in other short-term rates – those for bills, certificates of deposit, and in financial intermediaries' funding costs, and consequently the rates of interest at which they lend. Further out along the yield curve, of course, other factors impinge heavily – expectations of future inflation, expectations of future changes in monetary policy, and other demand and supply factors. But in Australia, intermediated finance and the bank bill market account for the vast bulk of debt financing. The cost of debt finance generally can, therefore, be influenced by monetary policy.
The next stage of the “transmission mechanism” in mind here is fairly conventional: there is substitutability between the various financial and real assets in the economy; savings, consumption and investment decisions are responsive to interest rates; goods prices and wages tend to be sticky, so that product and labour markets do not always clear instantaneously; monetary policy is non-neutral in the short run, but pretty much neutral in the long run. This is in accord with what Charles Freedman has termed “the mainstream central bank model” of recent years.[4] While several papers in this volume touch on the transmission process, none spell it out to any great extent. Several Conference participants felt that it was difficult to interpret the empirical papers in this volume without a clearer idea of the “model” that the authors had in mind. The following paragraphs attempt, in fairly non-technical language, to remedy that omission.
Suppose monetary policy is tightened. Following the rise in wholesale interest rates, the household sector will soon face higher interest rates as well. As a result, the cost of consuming today will rise relative to the cost of doing so in the future, which should discourage present consumption and encourage saving.[5] Investment spending by the household sector will also be affected. Residential construction in particular is usually quite responsive to interest rate changes. Traditionally, the mechanism was a credit-rationing one: market interest rates rose above the ceilings imposed on savings banks, the major providers of housing finance, and funds for investment in owner-occupied housing simply dried up. It is still too early to be sure about the deregulated (post-1986) regime, but it appears that increases in deregulated housing loan rates squeeze out the demand for funds by intending owner-occupiers, and that this can happen reasonably quickly.
Higher interest rates will also have an impact on businesses. The cost of debt finance rises with interest rates (though less so in after-tax terms). This may well come at a time when firms' needs for working capital rise, if sales fall and inventories unexpectedly rise, as might typically happen in the early stages of a business downturn. The cost of raising equity funds would also rise with the cost of debt finance, as shareholders demand higher returns on equity. If expected future profit opportunities are reduced, businesses' inclination to expand their activities will decline, and expenditure on real plant and equipment and construction will fall.
It is also important to note that much of monetary policy's impact depends on its effects on expectations. A rise in interest rates that is confidently expected to be short-lived may have little effect on spending and production. But if changes in monetary policy can substantially affect expectations about future interest rates, prices and output, decisions will be altered.
Other things given, the higher domestic interest rates will also be reflected in an exchange rate higher than would otherwise have been the case. This will, by the usual arguments, reduce the output of the tradeable goods sector. The evidence in the Stevens and Thorp paper is consistent with such an effect, though it says little about its size and speed. A separate question, though not one without interest, is the effect of monetary policy changes on the trade balance. On this, the Stevens and Thorp results suggest that there are two effects: if monetary policy slows down domestic spending, this of itself will tend to improve the trade balance, but in raising the exchange rate it will also have an effect (through relative prices) in the opposite direction. Clearly the relative size, speed and persistence of the two effects is crucial for a full discussion of this issue; unfortunately the technique used does not take us far down that path.
The above discussion has focussed on short-term responses of spending and output to changes in interest rates. The more important issue conceptually, is the effects of monetary policy on inflation. These effects can perhaps be thought of as operating through variations in spending and output compared to the “full employment” or “non-accelerating rate of inflation” level of output. While prices may initially be unresponsive to monetary policy changes, changes in the level of economic activity alter the relative bargaining strengths of sellers and buyers in both product and labour markets. The latter effect could occur either where there is centralised wage fixation, or where there is decentralised bargaining. Monetary policy can also work on inflationary expectations. To the extent that monetary policy actions have credibility, expectations will be adjusted, and the output effects of those actions correspondingly reduced.
Economists often express some discomfort with a representation of monetary policy developed around interest rates, because, on the surface at least, it fails to make explicit how nominal variables in the system are anchored. It is well established in theory, for example, that certain kinds of interest-rate setting behaviour make the system unstable and leave the price level indeterminate in the long run. The simplest instance of this problem would occur where the authorities set out to fix the nominal interest rate. Easing monetary policy to resist a tendency for nominal rates to rise, either in the face of a shift in inflationary expectations or because of some real sector shock which raises real interest rates, will mean that inflation will grow worse. That will strengthen the tendency for nominal rates to rise. If that is resisted by a further easing of monetary policy, the situation will get even worse, and so on. Inflation will go on rising; rather than stabilising the system in the face of shocks, monetary policy will reinforce the shocks.
Focusing exclusively on the real interest rate (if it can be measured) for policy purposes does not ensure price level stabilisation either. The “correct” real interest rate will stabilise the real economy, but real stabilisation can, in principle, be achieved at any price level or any rate of inflation.
Hence the arguments in favour of a quantitative intermediate objective for monetary policy in terms of a nominal aggregate like money or credit. In that case, the price level is tied down through the demand function for that nominal aggregate. An inflationary shock would show up as an increase in the nominal money stock. If the central bank were focusing on money, it would tighten policy, even though nominal interest rates would already be rising.
But what if, as has been suggested above for Australia, the demand function for money is unstable? Is there a way of conducting monetary policy which does not rely exclusively on money, but which avoids the possibility of prices not being tied down at all?
The Edey paper shows that while making interest rates the ultimate objective of monetary policy is a potentially disastrous strategy, a strategy of using interest rates as the instrument (or operational objective) in pursuit of a final objective such as the price level or nominal GDP can yield acceptable results. If interest rates are varied according to some function of the deviation of the final objective from its target path, the price level can be stabilised, provided the objective chosen is a nominal variable. Furthermore, if the demand for money is sufficiently unstable, this strategy will yield better results than a monetary target, in the sense of a lower variance of the target variable around the target path. This suggests that there is analytical sense in thinking about monetary policy in terms of short-term operational objectives for interest rates, against a background of a medium-term objective for nominal GDP or prices.
Thus this final paper in the present volume serves as a reminder that while both the conduct of monetary policy and the way it has its effects on the economy have changed over the past decade, the most important lesson of the 1970s is still valid: that while monetary policy can and does affect activity in the short run, its ultimate goal should be price stability.
Footnotes
Charles Goodhart, “The Conduct of Monetary Policy”, Economic Journal, 99, June 1989, pp 302–305. [1]
A selection of previous work is listed at the conclusion of this paper. [2]
Nor is it sufficient, in practice, simply to substitute the monetary base for “M”, and expect there to be a stable “money multiplier” as in the textbook. A given quantity of required reserves could be consistent with a multitude of total balance sheet sizes, depending on the extent to which banks used non-conventional deposits which have not, at least until recently, attracted reserve requirements. [3]
Charles Freedman, “Monetary Policy in the 1990s – Lessons and Challenges”, delivered to Federal Reserve Bank of Kansas City Symposium “Monetary Policy Issues in the 1990s”, August 1989, Symposium Volume forthcoming. [4]
There are income effects to be considered as well, of course. For debtors, the income effect reinforces the substitution effect. For creditors, the income effect works in the opposite direction: higher interest returns on existing savings increase consumption possibilities. The relative size of these effects is an empirical question. [5]
A Selection of Previous Reserve Bank Work on Money and Credit
Blundell-Wignall, A. and Thorp, S. (1987), “Money Demand, Own Interest Rates and Deregulation”, Reserve Bank of Australia Research Discussion Paper No. 8703, May.
Bullock, M., Stevens, G. and Thorp, S. (1988), “Do Financial Aggregates Lead Activity? A Preliminary Analysis”, Reserve Bank of Australia Research Discussion Paper No. 8803, January.
Jonson, P.D. (1987), “Monetary Indicators and the Economy”, Reserve Bank Bulletin, December.
Macfarlane, I.J. (1989), “Money, Credit and the Demand for Debt”, Reserve Bank Bulletin, May.
Macfarlane, I.J. (1989), “Policy Targets and Operating Procedures: The Australian Case”, presented to Federal Reserve Bank of Kansas City Symposium on “Monetary Policy Issues in the 1990s”, Symposium volume forthcoming.
Reserve Bank of Australia (1984), “Monetary Aggregates as Monetary Indicators”, Reserve Bank Bulletin, May.
Reserve Bank of Australia (1984/85), “Deregulation and Monetary Aggregates: Some Issues”, Reserve Bank Bulletin, December/January.
Reserve Bank of Australia (1987), “Measures of Financing”, Reserve Bank Bulletin, October.
Stevens, G., Thorp, S. and Anderson, J. (1987), “The Australian Demand Function for Money: Another Look at Stability”, Reserve Bank of Australia Research Discussion Paper No. 8701, January.
Veale, J.M., Boulton, L.F. and Tease, W.J. (1985), “The Demand for Money in Australia: A Selected Survey and Some Updated Results” in “Meeting on Monetary Issues”, Research Discussion Paper No 8502, November.