RDP 9703: The Implementation of Monetary Policy in Australia 3. The Thinking Underlying Current Arrangements

3.1 The Case for Announcements

Looking back, the Bank's view is that there have been significant net benefits in the practice of making announcements. Essentially these rest on the now well-documented arguments for central-bank accountability. Among the specific benefits:

  • The discipline of having to explain to the public the reasons for policy changes has led to greater rigour in the Bank's internal policy debates, and a clearer focus on the objective of the policy change.
  • Clear statements of the reasons for policy changes have led to greater community acceptance of monetary-policy decisions, and a better understanding in the community of the Bank's objectives. As a by-product, it has enhanced perceptions of the Bank's independence, as the community now more fully appreciates the Bank's responsibilities for monetary-policy decisions.
  • There are some signs that it has improved the transmission of monetary policy. There appears to be a more direct psychological impact on households and businesses, and the pass-through to bank interest rates is noticeably faster, particularly when monetary policy is being tightened. The reasons here are twofold. First, recognition lags – i.e. the time it took the market to discern a change in policy – have been eliminated. Second, there is more pressure on banks to respond to a publicly announced change in policy than was the case when policy changes were discernible only to professional market participants.

The process of announcing changes to the cash rate target has some of the same characteristics as a system of administered interest rates, and there was therefore some concern, when the Bank decided to make policy announcements, that it would reduce flexibility. In particular, it could lead to greater reticence in implementing policy changes and thereby increase response lags. In the event, however, this concern has proved unfounded.

Overall, the move by the Bank to announce and explain each policy change is seen, both in the Bank and in the community at large, as being a positive development and it would be difficult to envisage circumstances in which the Bank would now move away from this practice.

3.2 The Pattern of Interest-rate Changes

Between January 1990 and September 1996, there were 19 changes in monetary policy – 15 easings between 1990 to 1993 as the overnight interest rate was reduced from 18 per cent to 4.75 per cent; three tightenings in the second half of 1994; and an easing in July 1996. Of these, one involved a change in the cash rate of 150 basis points; eleven involved a change of 100 points; two involved a change of 75 points; and five involved a change of 50 points.

These changes in the overnight interest rate have generally been larger than is the case in many other developed countries; on average, the size of moves has tended to be about twice as large as those in other English-speaking countries (Table 1).

Table 1: International Comparison of Changes in Interest Rates
Tightening cycle
(1994)
Latest easing cycle
(1995 – September 1996)
Number of changes
 
Total change (basis points)
 
Average
change
(basis points)
Number of changes
 
Total change (basis points)
 
Average
change
(basis points)
Australia 3 275 92 1 50 50
US 7 300 43 3 75 25
Canada 10 450 45 17 425 25
UK 3 150 50 4 100 25

Despite the larger size of moves in Australia, interest rates have typically been adjusted multiple times in the same direction before a move is made in the other direction, a pattern which is common to most countries.[9] The Australian experience is summarised in Table 2. Using data on the target cash rate, we characterise each interest-rate change as a ‘continuation’ or a ‘reversal’, where a ‘continuation’ is a movement of the target rate in the same direction as the previous change, while a ‘reversal’ is a move in the opposite direction. The table shows that over the past decade the number of continuations is almost four times that of reversals. Over the period in which changes in the target rate have been announced, the ratio of continuations to reversals is even higher, reflecting the long series of reductions in interest rates in the early 1990s.

Table 2: Interest–rate Changes in Australia
Number of
changes
Average size
(basis points)
Average days since
previous change
January 1985 – September 1996
– Continuations 30 104 72
– Reversals 9 142 203
– Total 39 113 103
January 1990 – September 1996
– Continuations 16 89 88
– Reversals 3 75 409
– Total 19 87 138

Table 2 also shows the average size of the movements in the cash rate and the average number of days between movements. There is no systematic difference in the average size of reversals and continuations, but as the time lengthens since the previous policy adjustment, the probability that the next move will be a reversal appears to increase.[10]

The relatively large number of continuations suggests that interest-rate changes are positively autocorrelated. This implies that changes are predictable, in the sense that the probability of an interest rate increase and the probability of a decrease are not always equal. Any predictability of changes in the policy interest rate has been criticised by some economists who argue that the authorities should set the interest rate so that the probability that the next move is up is equal to the probability that it is down. This argument is typically based on a model which sees a very restricted role for monetary policy. For example, Barro (1989) argues that the monetary authority's task is simply to move the nominal interest rate in line with the equilibrium real rate, which is constantly changing in an unpredictable fashion. As a result, changes in the nominal rate should also be unpredictable.

In contrast, if the equilibrium real rate is constant, or varies around some (constant) average, movements in interest rates will sometimes be predictable. Suppose, for example, that the central bank raises nominal (and real) interest rates to combat an inflation shock. At some point in the future, one would expect that interest rates would need to be lowered, otherwise the real interest rate would be permanently higher. Given such an expectation, it can not be true that the interest rate is always set at the point where the expected change over any future period is zero. Put simply, if there is an equilibrium nominal interest rate and the current rate is not at the equilibrium, interest-rate changes will be predictable.

This predictability implies that interest rate moves are followed by reversals; it does not imply multiple movements of the policy interest rate in the same direction. Interest rates could be moved up in response to a shock and then moved back down in a single step to their initial value. This is the pattern that Goodhart (1996) has in mind when he argues that an important explanation for the persistence of inflation is that central banks ‘do not vary interest rates sufficiently aggressively, or promptly, to hold inflation to a desired path’. The argument is that by delaying interest-rate moves, or by moving by too small an amount, central banks are eventually forced to make further moves in the same direction. Goodhart argues that if a central bank was pursuing an inflation objective, the level of the official interest rate should be random (around some constant mean), so that only the next move (back to the mean) would be predictable.

In contrast, Blinder (1995, p.13) has implicitly argued that a sequence of predictable interest-rate movements represents optimal behaviour, stating that ‘a little stodginess at the central bank is entirely appropriate’. He suggests that central banks follow a strategy in which movements in interest rates are smaller than optimal, but are followed up with subsequent movements if things work out as expected.

There has been little formal justification for why central banks might act in this way.[11] In an attempt to fill some of the gaps, the remainder of this section is devoted to discussing possible explanations of the observed pattern of interest-rate changes (other than sub-optimal behaviour of the central bank). The most plausible explanations have their roots in the uncertainties that policy-makers face.

3.3 Reasons Why Central Banks Engage in Interest-rate Smoothing

There are three broad lines of argument why central banks may want to smooth interest-rate changes:[12]

  • slow adjustment of interest rates represents the optimal response to shocks, even when there are no costs of adjustment;
  • when there are adjustment costs, slow adjustment of interest rates reduces these costs; and
  • slow adjustment of interest rates is due to a combination of uncertainty and the costs incurred in changing the direction of interest rates.

3.3.1 Slow adjustment can be optimal

It is sometimes argued that the positive autocorrelation of interest-rate changes is prima facie evidence that central banks are not doing their job properly; that they are not moving aggressively enough to offset shocks. Certainly, one can build simple models which predict that a central bank with an inflation target will deliver uncorrelated interest-rate changes. However, slight variations to these models can deliver systematically positively autocorrelated interest-rate changes (Appendix B). The critical issue is whether the economy responds to a change in interest rates in the same way as it does to various other shocks. The model in Appendix B illustrates that if the response patterns are sufficiently different, moving interest rates multiple times in the same direction may represent the optimal policy response to a shock.

3.3.2 Adjustment costs

If interest rates are costly to adjust, it may be sensible for central banks to move gradually to their desired target. The critical point is that the costs must be increasing at an increasing rate in the size of the adjustment. If the costs are of this form, moving in small steps minimises the total adjustment costs. In contrast, if there is simply a fixed cost of adjusting interest rates, small adjustments will be ruled out, but so too would be multiple moves.

One reason why central banks avoid moving in one (large) step to the estimated new desired interest rate is that this could be very disruptive to financial markets, and could even threaten the stability of the financial system. In principle, a small change in interest rates which is accompanied by expectations of further moves should have a similar effect on long bond yields as a large, one-off move. However, in practice, it is questionable whether this is the case. Many financial prices react slowly to news. Recent experience suggests that even though a sequence of policy changes is expected, the full reaction in financial markets does not occur up front, but gradually over the period of adjustment of policy interest rates. Partly as a legacy of the history of relatively small movements, large up-front changes in policy interest rates risk causing much larger movements in securities prices. Even without this history, large unexpected changes in policy interest rates probably add to financial-market volatility, and in some cases concentrate the losses in particular institutions, perhaps threatening the stability of the financial system.

Similar issues arise for households and businesses. A gradual adjustment in policy interest rates can provide breathing space for borrowers to re-arrange their financial affairs, so that the liquidity effects of a change in interest rates are not as dramatic. This line of argument is obviously more relevant in financial systems in which variable-rate debt plays a significant role.

3.3.3 Uncertainty and costs of reversals

The most persuasive argument for moving gradually is that monetary policy is made under considerable uncertainty. Policy-makers do not know the true model of the economy, they do not know the exact impact that a change in interest rates will have on activity and inflation, and it can be difficult to assess the current state of the economy.

3.3.3.1 Uncertainty about the model

Blinder (1995) notes that uncertainty regarding the value of the parameters in the policy-makers' model justifies making smaller interest-rate movements than would otherwise be the case. The idea, based on earlier work by Brainard (1967), is that as the interest rate is moved further away from its average value, policy-makers become more uncertain about how a change in the interest rate will affect their objectives. This increased uncertainty reduces the optimal size of the policy move.

While the idea provides a convincing rationale for policy-makers being cautious in changing interest rates, it does not provide a complete explanation for multiple moves in the same direction. It could only do so if, after having moved the interest rate, the central bank learns some information which reduces the degree of parameter uncertainty. The gradual reduction in uncertainty would then lead to multiple moves towards the target rate. A difficulty with this argument is that the time between movements in policy interest rates is often too short for there to be any significant reduction in the degree of parameter uncertainty.

3.3.3.2 Uncertainty and costly reversals

Another model of the effects of uncertainty is suggested by the work stimulated by Dixit (1989).[13] In attempting to explain the slow evolution of the capital stock to its optimal level, Dixit proposed a model which considered the interactions of:

  • uncertainty about future returns;
  • the irreversibility of investment projects;[14] and
  • the possibility of delaying investment.

The analogy to monetary policy is clear:

  • policy-makers are uncertain about the future state of the economy;
  • changes in the direction of interest rates are seen as costly; and
  • there is always the possibility of delaying an interest-rate change until the policy-maker has more information.

The investment literature makes the point that the ability to delay investment decisions can profoundly affect the decision to invest today. If there are costs of reversing an investment, making a decision to invest today involves giving up the option of waiting for new information, and perhaps making a better decision tomorrow. As a result, it may be optimal to delay investment, or to make a smaller investment and subsequently increase its size if the uncertainty is resolved in the direction that was expected. While a delay in investment is a common prediction of these models, Abel et al. (1996) note that other outcomes are possible. For instance, if the cost of future investment is expected to rise, delaying an investment may be sub-optimal even in the presence of considerable uncertainty.

For these insights to apply to monetary policy, changing the direction of interest-rate movements must be costly – that is, there must be a cost to ‘reversing the investment’. The costs might arise in terms of additional volatility in financial markets and potential damage to the central bank's reputation.

To investigate the issue of whether changes in the direction of the cash rate lead to greater short-term volatility in security market yields we estimate the following equation,

where it is either the 90-day bank bill rate or the 10-year bond rate, Inline Equation is the target cash rate and REVERSAL is a dummy variable that takes a value of 1 if there is a reversal of policy on that day, and takes a value of zero otherwise. We use daily data and restrict the estimation period to the period over which announcements have been made. The results are presented in Table 3. If policy reversals cause increased volatility in the yields on 90-day bills or 10-year bonds the coefficient on the REVERSAL dummy variable should be significant and positive.

Table 3: Interest-rate Changes and Market Interest Rates: Regression Results
90-day bills 10-year bonds
Intercept (α) 3.2
(0.1)
6.5
(0.2)
Absolute change in target rate (β) 0.209
(0.013)
0.003
(0.017)
Dummy for reversal (δ) 6.2
(2.9)
14.7
(3.8)
Inline Equation 0.16 0.01
Number of observations 1,678 1,678
Notes: (a) Standard errors are in parentheses below coefficient estimates.
(b) Sample period is from 23 January 1990 to September 1996.
(c) Interest rates are expressed as number of basis points.

In both equations, a change in the direction of monetary policy appears to be associated with a larger than average movement in the market interest rate, with the effect being particularly pronounced for the 10-year bond yield. The estimates suggest that on days that policy is moved in the opposite direction from the previous move, bond yields have tended, on average, to move by an additional 15 basis points; for bill rates the figure is 6 basis points. This larger movement occurs despite the fact that the change in direction was not always unexpected. The results also confirm that the larger is the change in the cash rate, the larger the change in the bill rate. This suggests that large changes in policy interest rates would induce more volatility into the short end of the yield curve. There appears to be no such effect at the longer end of the curve. While it is difficult to draw strong conclusions due to the small number of reversals in the sample period, the results suggests that policy reversals increase short-term volatility across the yield curve. This issue is a topic for future research.

A more difficult issue to evaluate is whether this additional volatility is itself costly. Perhaps more difficult still is the issue of whether reversals would continue to generate additional volatility if they were more frequent. If reversals occurred as often as continuations, would they be less newsworthy and would they invoke a smaller reaction from the market? Or would the frequent reversals undermine confidence in the central bank, ultimately leading to greater volatility in financial markets?

The issue of how frequent reversals might affect the credibility of the central bank has importance beyond the implications for financial markets. If frequent changes in the direction of interest rates undermine public confidence in the central bank, inflation expectations might be higher than would otherwise be the case and the climate for investment may be adversely affected.

In Australia, as in all developed countries, changes in interest rates attract considerable public attention; on the day of the change, it is usually the leading story in the media. The Reserve Bank's practice of announcing and explaining interest-rate changes has tended to make policy changes more newsworthy, which may have made frequent changes in the direction of interest rates more difficult. In the absence of clearly defined shocks (such as a large exchange-rate change) it may be difficult for the central bank to make the case to the public that interest rates should be lowered one month, raised the next, and then lowered again a month later. Regardless of whether or not such a policy was ‘optimal’, the public would probably see the central bank as indecisive. As discussed earlier, the flexibility to change direction without adverse effects was an important reason why many central banks in the 1980s did not announce changes to policy.

While announcements might make reversals more difficult, they have significant benefits in terms of accountability and the transmission mechanism. These benefits currently outweigh any reduced flexibility that might be associated with the announcements. However, the need to minimise the probability of costly reversals means some caution is called for in moving rates. By making a smaller change, policy-makers can reduce the probability of having to make a costly reversal. If developments turn out as expected, policy can be moved again in the same direction; if the unexpected happens and it turns out that no change in policy was required, the costs of reversing policy are saved. The end result of waiting for uncertainty to be partly resolved is systematically positively autocorrelated interest-rate movements and positively autocorrelated inflation and economic activity.

The major qualification to this line of argument arises if the size of the required policy adjustment depends upon the speed and magnitude of the initial policy adjustment. Suppose a delay in adjusting policy to an inflationary shock led to inflation expectations rising. The end result would be much higher interest rates, as the central bank struggled to reverse the rise in expectations of future inflation. In this case, an early and large increase in interest rates may be warranted, despite the fact that an increase of this type increased the probability that the policy move would have to be reversed.

Finally, the stronger is a central bank's inflation record, the longer may be the period it can wait before adjusting policy. If the central bank has high credibility, inflation expectations are less likely to adjust up, and this should give the authorities more scope to wait and to assess the exact nature of the shock and the appropriate policy response. In so doing, increased credibility might provide more scope for the central bank to avoid costly reversals of policy.

3.3.3.3 Uncertainty and the decision-making process

A third way that uncertainty contributes to the positive autocorrelation of interest-rate changes is the impact that it has on the decision-making process. There is rarely unanimity among experts concerning the exact magnitude of necessary monetary-policy adjustments. Many of the differences of opinion arise from the uncertainty regarding the current state of the economy, the exact nature of shocks and the exact structure of the economy and the associated policy multipliers. Decision making by committee creates an environment in which the various views can be discussed and reduces the probability that extreme decisions are made. However, committee-based decision making can also systematically affect the way policy responds to shocks. It can lead to compromise solutions which involve a smaller initial change in interest rates followed up by a further increase, when the case becomes more compelling.

A related issue is that monetary policy needs to be broadly acceptable to the community as a whole. Again, because of uncertainty about the future, it may be difficult to convince the community that a large interest-rate change is the appropriate response to a shock. The case for such a change is likely to depend heavily on the central bank's forecasts for activity and inflation. These forecasts might differ from those of other forecasters and the public may not be convinced that the large change in interest rates is required. With little history of large one-off interest-rate movements, the public in most countries are unlikely to find such movements acceptable. The ability of the central bank to convince the public of the need for large movements in interest rates will depend importantly on the confidence that the central bank enjoys in the community. A record of good policy-making would undoubtedly assist in this regard. But, even central banks with good records must constantly weigh up the risks of policy errors as public support can quickly be lost. As a result, to maintain a broad consensus concerning the direction of monetary policy, interest-rate changes may be smaller and more gradual than would have otherwise been the case.

3.4 An Assessment

Three of the defining characteristics of Australia's recent monetary policy arrangements are:

  • changes in policy interest rates have been explicit and the reasons for the change are explained in detail;
  • individual changes in policy interest rates have tended to be large by international standards; and
  • policy interest rates have typically been moved multiple times in one direction before a move in the other direction is made.

As discussed above, clear announcements are not only important from the perspective of accountability, but they also mean that interest-rate changes more effectively influence people's expectations about future economic and financial conditions. In turn, the monetary transmission process should be quicker and the amplitude of the interest-rate cycle smaller.

Beneficial announcement effects are more likely to be achieved if the changes in policy interest rates retain their newsworthiness. If changes in rates occur extremely frequently, or are very small, their newsworthiness is likely to be diminished, and they are less likely to force a relatively quick change in banks' posted lending rates. Changes in these rates reinforce the attention that the change in official interest rates receives. If policy interest rates were changed frequently, and/or by small amounts, the immediate link between monetary policy and lending rates might be weakened, reducing the overall effectiveness of the policy changes.

There is, however, no clear, unequivocal basis for deciding what is the optimal size and pattern of interest-rate adjustments. The decision is influenced by the expected overall move in rates that is required; by the impact that changes of various sizes will have on people's expectations; by the costs that are paid if the policy change has to be reversed; and by the degree of uncertainty. In turn, these factors will be influenced by the nature of the forces necessitating the policy change, the institutional structure of the economy and financial system, and the historical context of policy changes.

The pattern of interest-rate changes that has been adopted by the RBA is one that has delivered positively autocorrelated inflation rates. The same is true in almost all countries. While the simple models of some economists suggest that this persistence could be eliminated by changing the way in which central banks move their policy instrument, this has rarely been attempted, or achieved, in the real world. The range of uncertainties that policy-makers face provide a solid explanation for the persistence of interest rates, and ultimately inflation rates. It is a question of attempting to balance the beneficial announcement effects of relatively large changes in interest rates against the costs that can arise if policy changes are frequently reversed.

Footnotes

See for example Goodhart (1996) and Rudebusch (1995). [9]

Rudebusch's (1995) analysis of changes in the Federal Funds rate between 1974 and 1992 suggests that in the first couple of weeks after a change there is a higher probability that the next move will be in the same direction, rather than in the opposite direction. After four weeks the probabilities are broadly equal. [10]

There is, however, an extensive literature on the issue of whether monetary policy should smooth interest rates in the face of changes in money demand, with Poole (1970) being the seminal article. For example, Mankiw and Miron (1991) have examined the issue of whether central banks should smooth interest rates in the face of seasonal changes in the demand for money. Mankiw (1987) and Barro (1989) have also argued that maintaining a smooth interest rate may be optimal from the point of view of smoothing the inflation tax. [11]

The issue of interest-rate smoothing is more fully discussed in (Lowe and Ellis 1997). [12]

For a comprehensive review see Dixit and Pindyck (1994). [13]

More recent models have allowed investment to be reversed, but at a cost (Abel et al. 1996). [14]