RDP 9604: Issues in Modelling Monetary Policy 2. Modelling the Policy Framework

Essentials of the policy framework that need to be understood for the purposes of model-building can be briefly outlined in terms of instruments, objectives and decision criteria.

The instrument of monetary policy is the short-term interest rate – really, the overnight cash rate although, for the purposes of a quarterly or annual econometric model, this might be regarded as largely interchangeable with the 90-day bill rate. Since January 1990, changes to the desired cash rate have been publicly announced, a system that has been conducive to keeping cash rates in a tight range around the desired level (Figure 1). Cash-rate announcements have been accompanied on each occasion by public statements setting out the reasons for the change.

The objectives of monetary policy in terms of inflation and output performance have been described in detail in a number of public statements and publications by the Bank over recent years^[2], and were formally set out in the recent Statement on the Conduct of Monetary Policy^[3]. That statement notes that the policy objectives defined in the Reserve Bank Act ‘allow the Reserve Bank to focus on price (currency) stability while taking account of the implications of monetary policy for activity and, therefore, employment in the short term’. The Statement also sets out the Bank's inflation objective in numerical terms:

‘In pursuing the goal of medium term price stability the Reserve Bank has adopted the objective of keeping underlying inflation between 2 and 3 per cent, on average, over the cycle. This formulation allows for the natural short run variation in underlying inflation over the cycle while preserving a clearly identifiable benchmark performance over time.’

A number of points can be emphasised concerning the implications of this formulation. These concern both the inflation objective itself, and the relationship between objectives for inflation and output. With regard to the inflation objective, it should be noted that the specified 2–3 per cent range does not represent a hard-edged target band within which the inflation rate is to be confined in every period. The range has instead been described as a ‘thick point’, an indicator of central tendency rather than a target band, and the use of a range rather than a point is intended to recognise the imprecision with which inflation can be controlled and to avoid the spurious accuracy of point targets. Also important is that the objective is to be achieved on average over time rather than in every period, so it is a medium-term rather than fixed-period objective, where ‘medium-term’ is broadly defined as the length of a business cycle. The inflation objective is specified in underlying terms, using a measure of prices (the Treasury's underlying rate) which abstracts from the effects on the CPI of erratic factors such as government charges and interest-rate movements. These features of the inflation objective leave room for flexibility for policy to take into account short-run developments in output and employment and to consider both the real economy and inflationary short-run consequences of shocks when formulating policy responses. This sort of flexibility is in fact built in, in varying degrees, to the systems of other inflation-targeting countries. Over longer periods when monetary policy cannot be expected to have lasting effects on output and employment, the policy objective anchors inflation in the 2–3 per cent range.^[4]

Given the instrument and objective of policy, it remains to comment on the decision criteria linking policy decisions to information about the objective variables. Here two points can be made. First, because of information and impact lags, the objective requires monetary policy to be explicitly forward-looking: thus the policy instrument will be adjusted in response to expectations of price and output variables in a way consistent with keeping inflation on the desired medium-term path. This will in practice mean looking at a range of variables that provide useful forward information about price and output trends. This is not a ‘look at everything’ rule but a criterion for interpreting information from a potentially large number of sources. Second, there is no formal commitment to a policy role for any particular intermediate variables such as monetary or financial aggregates. These and other variables are evaluated pragmatically for their usefulness as forward indicators and are in no sense to be regarded as ‘intermediate objectives’. In practice the monetary aggregates in Australia have been highly unstable in their behaviour over at least the past decade or so and their information content is regarded as small.^[5] An implication of this discussion is that there is no obvious definition of an ‘unchanged policy’ stance in terms of a constant interest rate or money-supply growth rate. In a modelling context, ‘unchanged policy’ needs to be defined in terms of a given decision rule relating the short-term interest rate to information about prices and output in a way consistent with achievement of the medium-term objective.

In qualitative terms the conceptual framework outlined above is broadly consistent with those operating in a variety of countries, including those with formally legislated inflation targets and others with somewhat broader objectives. In a recent survey Haldane (1995) argues that the basic features of this framework – ie the use of an interest rate instrument to directly target final objectives for output and inflation over a period of time – are applicable to a wide range of countries, with cross-country differences arising mainly on more secondary matters such as time frames, tolerance bands and the like. This role for interest rates is also increasingly recognised in the academic literature, as for example in studies by Bernanke and Blinder (1990), Taylor (1993) and Fuhrer and Moore (1995) which empirically represent US monetary policy decisions by changes in the federal funds rate. Blinder (1996) provides a neat summary of this view which is worth quoting in detail:

Today, the real short-term interest rate is the logical replacement for the money growth rate. Its ‘neutral’ value can be estimated from history and/or from econometric models, although it can never be known with certainty. ‘Tight’ monetary policy can then be defined as keeping the real interest rate higher than neutral. Such a policy can be expected to contract the economy, after a lag, and reduce inflation, after an even longer lag.

There has been much talk in recent years of the problem posed by the loss of the so-called nominal anchor for monetary policy when monetary aggregates are bandoned in favour of interest rates. The new ‘anchor’ I am tacitly proposing has three pieces:

1. the central bank's long-run inflation target;
2. its commitment to keep real interest rates higher than neutral when inflation is above target, other things equal;
3. understanding that nominal interest rates must not be held fixed, but must be adjusted for inflation because it is real interest rates that matter.

Despite these trends, one gets the impression that it is still not quite academically respectable to replace money supplies with interest rates in formal models.^[6] This appears to stem from theoretical results such as reported by Sargent (1979) to the effect that exogenous interest-rate-setting policies are unstable or leave prices indeterminate.^[7] It should be noted, however, that these results apply only under the extreme assumption that interest-rate setting is unrelated to nominal policy objectives. McCallum (1986) and others have shown that well-specified rate-setting policies are fully capable of providing a suitable nominal anchor, provided the rate-setting rule is specified as a function that includes a nominal target variable, a requirement that would clearly be met by inflation targeting systems or by a range of other possible systems such as nominal income targets.

Full recognition of the role of interest rates as the policy instrument in a modelling context cannot be achieved simply by inverting money-demand functions to produce what might be interpreted as rate-setting policy rules. This procedure does not really get around the problem of identifying a given policy setting with an exogenous money-growth path. To illustrate, suppose a simplified money-demand function of the following form (ignoring dynamics):

where variables have the obvious definitions and are measured relative to steady-state values. This can be inverted to give an expression for the interest rate conditional on the other variables:

For the purposes of policy simulation it is possible to think of an exogenous money-growth path in this equation as being akin to a kind of target index for an appropriately weighted average of prices and output, so the approach can be argued to have some validity for policy analysis irrespective of the poor historical performance of monetary aggregates as policy indicators. On this interpretation the interest-rate equation would just represent a reaction function to deviations of the price and output index from target. Since this is broadly what has been advocated in the preceding discussion, it might be thought that this kind of policy assumption would yield reasonable results. Our point however is that the parameters and dynamics of the inverted money-demand function do not necessarily have anything to do with those of a sensible policy-reaction function, except to the extent that it is legitimate to think of the monetary authority as adhering to a strict money target.^[8] An alternative way of making this point would be to add to the above equations a money-supply function that smooths the interest rate, say:

This can in principle be combined with equation (1) to eliminate money and arrive at a policy reaction function whose parameters would differ from those of equation (2). If we wish to analyse policy behaviour under reasonable assumptions about the policy framework, there is thus no reason to think it would look like a variant of equation (2) with exogenous money.

In our view a more promising approach is found in recent studies by Bryant, Mann and Hooper (1993) and by Taylor (1993, 1995), which bypass the money-demand function altogether and specify interest-rate policies as simple functions of price and output variables with simple dynamics. This approach has a number of advantages in that it puts the focus on comparative stabilising properties of alternative rules, within a class of rules that has appealing properties, rather than on the dynamics of responses to shocks under any one particular rule.^[9] Taylor (1993) obtains the useful result that interest-rate setting in the US since the mid 1980s is well approximated by simple functions of price and output movements, and his empirical model suggests that this class of policy rules is capable of approximating the stabilisation properties of an optimal rule.

Footnotes

See in particular Fraser (1994), Debelle and Stevens (1995), Grenville (1996). [2]

Full details are set out in ‘Statement on the Conduct of Monetary Policy’ (1996). [3]

It should be noted in passing that there is nothing illogical about having both prices and output in the objective function. This kind of formulation is a standard feature of literature on the theory of monetary policy, as for example in the textbook treatment of Blanchard and Fischer (1989). [4]

See de Brouwer, Ng and Subbaraman (1993), and also the discussion in Section 3.2 below. [5]

For example, Mishkin's (1995) summary of a recent symposium on the monetary transmission mechanism repeatedly defines a monetary policy shock as ‘M↓’. [6]

In the simplest case, that of a fixed nominal interest rate, inflationary shocks reduce the real interest rate and are therefore self-reinforcing. [7]

Edey (1990) presents a general argument that policy under an interest-rate reaction function with freely-chosen parameters dominates money targeting in terms of stabilisation of prices and output. [8]

Edey (1989, 1990) gives a discussion of theoretical principles that might guide the design of simple rate-setting policy rules with desirable stabilisation properties. [9]