RBA Annual Conference – 1989 Conference on Money and Credit: Summary of Discussion
Monetary Targeting: The International Experience
The discussants agreed with the paper's conclusions that central banks began targeting monetary aggregates as a tactical response to a specific set of economic conditions, not as a revolutionary change to the policy-making process.
Some participants felt that in several respects, monetary targeting had been unsuccessful. The paper had noted, for example, that there was a very strong tendency for targets to be missed, usually by overrunning the desired rate of monetary growth. In addition, lasting progress on inflation was not made until the mid 1980s, long after targets were first introduced. On the other hand, others viewed this differently: they felt that it was possible to argue that targeting had been abandoned or downgraded in importance only after it had helped to achieve some success against inflation.
It was noted that for targeting to be a “good” policy, one condition is an appropriate choice of aggregate to target. In some countries, the authorities met the monetary target for the chosen aggregate, but monetary policy was still not very effective: intermediaries switched business towards unrestricted products, or developed alternative ways to avoid controls. In this context, the authors reiterated a point made in the paper: rather than simply comparing targets and outcomes, it is important to evaluate targeting, and monetary policy generally, on the basis of the behaviour of the ultimate objectives of policy, such as prices or nominal income. In the case of Canada, for example, targets had generally been achieved, but there had been little success in reducing inflation until after targeting was abandoned.
It was noted that technical change, and the effects of deregulation and innovation, had made targeting a very difficult strategy to follow in a number of cases. This had led to the abandonment or down-grading of targets. The discussion took up the question of whether we might expect to see an early return to targeting in Australia once the effects of deregulation have settled down. The majority of participants thought that this was unlikely.
A major reason offered was that even if the money/income relationship stabilised, it would be some years in practice before stability could be clearly recognised and a targeting policy be implemented with confidence. For the moment, then, it was suggested that the best that can be hoped for aggregates is that their broad trends will serve as useful intermediate indicators of the economy's general direction. Of the available aggregates, credit may conceptually be the best measure, because credit covers the effects of offshore (and other types of non-deposit) funding, although the credit/GDP relationship has shifted in the 1980s, and credit as conventionally measured excludes direct foreign lending.
At a more conceptual level, some participants ventured that it was unlikely that a stable money/income relationship would re-emerge at all. This argument was based on the view that the earlier apparent stability of the money/income relationship was itself a product of the then-existing regulations. With those regulations removed, and with financial innovation likely to continue, ongoing instability of the money/income relationship was the most probable outcome, ruling out monetary targeting as a viable strategy.
The Relationship Between Financial Indicators and Economic Activity: 1968–87
Results presented on the relationship between money and credit aggregates and activity were generally accepted. None of the participants wished to put the view that the relationships were more stable than implied by the paper. There was, however, a fair amount of discussion on the relationship between interest rates and economic activity.
Several participants said that they would like to have seen an account of the behavioural model that lay behind the comparison of interest rates with real and nominal activity. The authors conceded that the paper did not explicitly spell out a transmission mechanism, though parts of it had been given in other places.[1] The aim of this paper was really to investigate, in a fairly straightforward fashion, whether there were any reliable “stylised facts”.
Some participants felt that too much emphasis had been given to nominal interest rates, and that the work should be broadened to look at real interest rates. They emphasised the danger (which had been noted in the paper) of relying on any apparent relationship between nominal interest rates and real activity in a period where inflationary expectations were changing. In response, the authors pointed out that the emphasis had been placed on a nominal interest rate largely because this was the day-to-day operating instrument that the central bank had at its disposal. Nevertheless, they conceded that there was a good case to repeat some of the analysis using real interest rates, even though there can be lengthy arguments over the correct calculation of a real interest rate, and whether any desired real interest rate can actually be achieved as an operating objective.
Some participants argued that there was a good case for using a measure of the yield curve as a proxy for the tightness of financial conditions. The authors saw some value in this suggestion, and felt that it was related to the point about real interest rates. They suggested that changes in real interest rates and changes in the long-short spread were quite closely related over the 1980s. (Over the 1970s, on the other hand, there would have been a greater difference between the two indicators.)[2]
Other participants suggested that, as well as examining the relationship between financial indicators and real or nominal economic activity, the authors should take steps towards incorporating analysis on the relationship between those variables and the rate of inflation.
A different line of discussion was pursued by a number of other participants who drew out the limitations of trying to reach conclusions using only simple techniques, such as graphs or correlation coefficients. They felt that there were a number of influences on the cycle in economic activity and inflation that came from totally non-monetary sources; for example, the two oil price shocks of the 1970s, the wage explosions in the mid 1970s and early 1980s, and the sharp movement in the terms of trade in the mid 1980s. These could be expected to have a major influence, and since a number of those shocks have been correlated with movements in interest rates, their effects may show up in the bivariate relationships in the study.
The authors conceded that this would be a valid methological criticism of any study using simple techniques. Only a fully-articulated econometric model would not suffer from this problem, but that would be a different approach altogether (and would have advantages and disadvantages of its own), and a much larger undertaking than the present exercise.
A final suggestion was that future research might look at the effect of financial variables on particular components of final demand; for example, business fixed investment, investment in dwellings, consumption, stockbuilding, etc. The authors noted that some research along these lines had been done earlier in this project but the results had not been incorporated in the final papers. The representatives from Treasury noted that such work had been undertaken recently in Treasury.[3]
The Relationship between Financial Indicators and Economic Activity: Some Further Evidence
This paper was presented as an adjunct to that above. It used more rigorous techniques to examine the same set of questions about leads and lags.
Several participants were interested to know why some results of the earlier paper were not confirmed by these further tests, and in particular, why no statistically significant leading relationship from interest rates to activity could be found.
In response, the authors suggested that the tests employed in this paper are much more difficult to pass than those used in the earlier paper. They also conceded that it might be too much to expect to find unidirectional “causation” from interest rates to activity. If monetary policy reacts to activity, as seemed likely, then activity affects interest rates, as well as interest rates affecting activity.
One participant suggested that drawing stylised facts from Vector Auto-Regression (VAR) models of this type presents some difficulties, in the sense that the results are model-dependent, and the techniques rely on the inclusion of all relevant variables in the models. Earlier studies using VAR techniques had found that the results can be sensitive to the inclusion or exclusion of particular variables. The authors noted that they had this concern themselves. Hence the fairly large number of combinations of variables which were used: as much care as possible had been taken to prevent conclusions being drawn on the basis of only a small subset of possible results.
A number of participants felt that it is difficult, using these models, to take full account of the changing institutional environment, variations in the prevailing policy regime (including the period of monetary targeting and its subsequent abandonment) and other shocks to the economy; others noted that external shocks were covered, at least partially, where trade-sector variables were included in the models, and also that some shocks may be captured by the inclusion of lagged values of GDP.
Changes in the Behaviour of Banks and their Implications for Financial Aggregates
The discussants noted that this paper complemented the work presented in the two previous papers, by explaining some of the recent changes in the behaviour of money and credit aggregates.
Changes in the pattern of banks' holdings of excess LGS assets were largely the result of the deregulation of interest rates, which allowed the banks to implement strategies of liability-management. One discussant suggested that the smoothing of the pattern of tax payments over the year and insulation against capital flows provided by the floating exchange rate had also contributed.
A large part of the discussion focussed on the distortions caused by the SRD arrangements and the consequent growth of financing that was not funded by conventional deposits. As a result, there was much discussion about the bill market and its prospects.
Prior to the recent changes to SRD requirements, bank-bill funding had some clear advantages over ordinary loans. Bills offered customers fixed-rate as opposed to floating-rate facilities, and the competitiveness of the bill market drove down margins and the cost of finance. From the banks' point of view, bill lines had effectively circumvented quantitative lending guidelines, and the securitisation process (by which bills can be on-sold in the market) reduced risk and required less capital backing; it had thus allowed banks to expand their businesses more quickly.
Participants noted that the attractions for borrowers of bill finance compared to conventional loans had declined after the September 1988 change to the SRD arrangements. Some participants expressed surprise, however, at the extent of the shift back to conventional deposits apparently arising from the change, since funds from the SRD accounts were to be released, and therefore banks' average cost of funds lowered, only gradually over a long period. In response, it was suggested that the new arrangements substantially lowered the marginal cost of deposits immediately, and that this explained, at least to some extent, why the growth of deposits had been so fast recently. It was agreed that the portfolio adjustments induced by the changed SRD arrangements were likely to continue for some time.
The recent rapid overall growth in credit (linked to the expansion of the bill market) was discussed at some length. A number of participants noted that the growth of credit indicated that there has been substantial expansion of the community's aggregate balance sheet. The paper had covered changes related to the removal of particular regulations. Other factors, beyond the scope of the paper, such as behavioural changes related to technical change or internationalisation of the financial system, have also been important. All of these factors have implications for the way the financial system works, and the way monetary policy has its effect.
Some participants put forward the view that growth in corporate treasury activity, with increases in both sides of balance sheets, was a contributing factor in the growth of credit. Others noted, however, that preliminary research had been unable to uncover much hard evidence that growth in corporate treasury operations, if it had occurred, was large enough to account for very much of the growth in credit over recent years.
One set of regulatory changes not discussed by the paper was the new capital adequacy requirements. These changes are likely, in future, to increase the costs of some types of bank business. Large corporations which consequently face higher bank charges may turn to non-banks for funding or other services, and/or may look more closely at borrowing by issuing their own paper in the market.
At the same time, the new capital requirements would encourage banks to increasingly conduct their business on a group basis. This diversity of organisations under one umbrella would to some extent make a focus on the “banking” arm itself redundant within the new regulatory setting.
Monetary Policy Instruments: A Theoretical Analysis
Discussants expressed general agreement that the paper produced intuitively plausible results concerning the use of interest rates as an operating instrument of monetary policy. Specifically, it was agreed that an interest rate instrument could, under certain conditions, tie down the price level, and that a simple operating rule for interest rates would be superior to a simple monetary control rule if the demand for money were sufficiently unstable.
The discussion focussed on three main issues. The first was the question of whether simple or complicated decision rules are preferable. It was pointed out in discussion that optimal policy rules generally make use of all available information, whereas the policy rules analysed in the paper were arbitrarily restricted to be of a very simple functional form. Specifically, interest rates were assumed to be adjusted in response to movements in only a single target variable, and the paper selected only a small class of possible target variables for consideration. Quite clearly, if a wider choice of policy rules were assumed to be available to policy makers, it would be possible to find better rules than those discussed in the paper. The author agreed with this point in principle, but argued that the aim of the analysis was not to find a general solution for the optimal rule, which would inevitably be complicated and model-specific, but to consider the relative merits of a number of simple rules which may have some practical appeal. In favour of simple rules it was argued that complexity can only be justified when the authorities have extensive knowledge of the true model of the economy, whereas in reality such knowledge of the “true” model is rather sketchy. This is an argument often used by monetarists in favour of simple money growth rules, and it can correspondingly be used to defend the idea of simple interest rate rules. The aim is to find a rule which has desirable stabilisation properties across a range of models, and which can be implemented without detailed knowledge of the true model structure. A second possible advantage of a simple rule is that it encourages accountability, because it is easier to evaluate the authorities' performance against a simple standard than against a complicated one.
As a related point, it was suggested that a more rigorous way of restricting the class of rules available to the authorities would be to make specific assumptions about the information available, and then assume that policy makes optimal use of that information. This would also allow more general questions concerning the role of differences in information possessed by public and private agents in the model. Other participants doubted that significant information differences exist in practice; rather, the informed public has access to much the same information as policymakers via statistical publications and the financial press.
The second main issue of discussion concerned the definition of money in the model. This question was left open in the paper, but the author suggested in discussion that a money base definition would be most consistent with the theoretical analysis, under the presumption that it could be directly controlled by the authorities. Choice of the definition of money was not fundamental to the analysis, however, since money played no role in the model solutions under the interest rate rules considered in the paper. The important point is that simple interest rate rules dominate simple money supply rules unless some monetary aggregate can be found for which the demand function is sufficiently stable. Most participants felt that the money base was not a very useful focus for policy in Australia, and some doubted whether it could be effectively controlled by the authorities.
A third focus of discussion was on the topic of base drift, which refers to the practice of taking actual outcomes as the base from which targets are calculated in each new period. This is in contrast to a policy without base drift, which would seek to correct for any past failures to hit targets. The paper had argued that in a model with nominal neutrality and rational expectations, policies with and without base drift would produce identical time paths for output. The only difference in outcomes would be that under base drift, the unconditional variance of the price level would become infinite, but this would have no adverse consequences for the real economy. There would therefore be no strong grounds for preferring one form of policy over the other. Introducing a simple form of non-neutrality into the model, however, has the effect of improving the attractiveness of allowing base drift, because this avoids the implied negative effect on output when the authorities try to correct for past failures to hit targets. Assuming that the unconditional variance of prices is a relatively unimportant objective, this could therefore be put forward as an argument in favour of base drift.
Discussants disputed this conclusion on two grounds. First, it was suggested that the question of base drift could have been more formally treated in the optimising framework used in the paper, by explicitly allowing the authorities to choose some optimal degree of drift. This would be more general than assuming that there was a straight choice between full drift and zero drift. Secondly, it was argued that in a more general and more realistic model containing some nominal rigidities, the unconditional variance of the price level may become an important policy objective. Long-run variations in the price level would cause persistent distortions in relative prices or real wages, which may have substantial costs for the real economy.
Summing Up Discussion
At the conclusion of the formal papers, there was a general discussion, drawing together the themes of the various papers.
One of the major themes of the conference had been the difficulties in using financial aggregates: broader aggregates tended to be lagging indicators, and some had been subject to shifts in their relationships with national income or spending. Narrower aggregates were very volatile, and intermediate aggregates had been subject to instability. These observations were broadly accepted.
At the same time, there had been a marked change in the way people thought about the effects of monetary policy on the economy. One view put was that the primacy of “money”, which had characterised debate in the 1970s, had been replaced by a view in which monetary policy operated through, first, the effects of open market operations on asset prices and rates of return on financial assets, and secondly, through the response of spending and production to those financial changes. Financial assets and liabilities had become more substitutable; a host of new instruments had evolved; risk management techniques had become more sophisticated; average holdings of financial assets and liabilities had become much larger. In short, it was suggested that this world is better (though perhaps not perfectly) described by the “Tobin framework”, than by the “Friedman framework”.
The fundamental objectives of monetary policy, however, had not changed: most held that an acceptable long-run performance on prices was the best ultimate objective for monetary policy. While the research presented had suggested that monetary policy may have effects on real output and spending in the short term, it was noted that monetary policy's primary goal should be inflation or nominal income, and that any shorter-term “real” objectives had to be seen in that context.
On inflation, there was some discussion about the costs to output of reducing inflation further. Some questioned whether these were worth bearing. On the other hand, it was noted that there is a substantial cost to be born in accepting the present level of inflation, because of the interaction of inflation with the tax system and the efficiency losses caused by that distortion (notwithstanding the fact that the introduction of dividend imputation in Australia had reduced these sorts of distortions to some degree).
This highlighted the need for a better understanding of how output and prices interact. It was suggested that this was an important topic for future research. Some participants also suggested that the particular prices captured by the measure of inflation used in policy analysis are also important. At present, economists tend to focus on changes in the prices of current goods and services; little account is taken of changes in asset prices. But asset prices may be quite important, since they enter consumption and investment decisions through valuation effects on wealth. Such price changes had been an important feature of the 1980s.
Footnotes
See, for example, I.J. Macfarlane “Methods of Monetary Control in Australia” (1984), reprinted in D.J. Juttner and T. Valentine (eds.) The Economics and Management of Financial Institutions, Longman Cheshire, 1987 and “The Reserve Bank's Domestic Market Operations”, Reserve Bank of Australia Bulletin, June 1985. [1]
Follow up work looking at real interest rates and the yield curve is presented elsewhere in this volume. [2]
D.J. Bassanese and G. Debelle, “The Relationship Between Short-Term Interest Rates and Components of Economic Activity 1968–1988”, Treasury Seminar Paper, November 1988. [3]