6  Inflation targeting and the output/inflation variability trade-off

This section examines how monetary policy has affected the relationship between output variability and inflation variability before and during inflation targeting. Evaluations of monetary policy reaction functions using theoretical and calibrated model simulations typically assume the central bank’s objective is to minimise a weighted sum of the squared deviations of inflation and real output from their target values. The target for real output is typically the natural rate while the target for the rate of inflation is typically assumed to be zero. The relative weights assigned to the inflation and output targets vary according to central bank preferences and characteristics of the economy. Expression (16) is an example of the type of central bank objective function typically assumed in the literature.

(16)    

The object of monetary policy is assumed to be the minimisation of the loss function represented by expression (16). The parameter α is the relative weight on output deviations. The target variables are assumed to be potential output (so that y is the deviation of output from trend) and zero trend inflation (so that π is actual inflation).

Clarida, Gali and Gertler (1999) discuss the theoretical implications and Hunt and Orr (1998) and Drew and Orr (1999) provide examples of calibrated model simulation based derivations of optimal monetary policy rules using this type of objective function. In the context of dynamic general equilibrium models with money and temporary nominal rigidities, the policy problem is to choose a time path for the nominal interest rate to influence time paths for the target variables that maximise the objective function subject to the constraints of the private sector and behavioural functions of other policy institutions. These target variables depend not only on current policy but also on expectations about future policy. Therefore credibility of future policy intentions can influence the cost of lower inflation in terms of foregone output during the transition to low inflation. In this framework, an important result to emerge is that, to the extent cost-push inflation is present, there exists a short-run trade-off between inflation and output variability around the target values (Taylor, 1979).

The importance of this trade-off rests on the assumption that the objective function is maximized by minimising the deviations of actual from potential output and actual inflation from target inflation. Some formal justification for this is provided by Rotemberg and Woodford (1999). Inflation distorts intertemporal decisions with respect to labour supply and consumption. In the presence of nominal rigidities, shocks can force output to vary from the optimal or the natural output level. Rotemberg and Woodford show that the weights assigned to inflation and output deviations in the central bank loss function can be derived from a utility function that includes output and inflation variances.

An implication of this type of framework is that if the central bank is to maximise its objective function it must aim for convergence of inflation to its target over time. In general this would imply a gradual adjustment taking into account the implications for output variability. The speed at which central banks decide to pursue inflation targets will determine whether its procedure is characterised as strict inflation targeting, flexible inflation targeting or strict output targeting (Svensson, 2001). Strict inflation targeting, involving adjustment to target inflation at a faster rate, would only be optimal, for example, if there was no cost inflation or if there was no concern for output deviations from the target rate of output. It could also occur if for some reason the central bank had a higher preference for minimising inflation deviations and a lower preference for minimising output deviations than society.

This framework offers other important insights to guide monetary policy. These include the optimal response of monetary policy to demand and supply shocks and how monetary policy responses should be adapted in the presence of various forms of uncertainty, and the implications of improved monetary policy credibility and inflation inertia on the trade-off. In general, if price-setting depends on expectations of future economic conditions, improvements to policy credibility will improve the trade-off between inflation and output thereby enabling the central bank to converge on the inflation target with less cost in terms of output variability.[17] Also relevant to our analysis of the New Zealand experience are the implications of inflation inertia. In the presence of menu costs of price adjustment, Ball, Mankiw and Romer (1988) show that lower rates of trend inflation tend to induce inflation inertia and increase the output costs of lowering inflation.

Assuming the target inflation rate for the Reserve Bank is trend inflation, the inflation variability component in a central bank loss function of the type illustrated by expression (16), and drawing on expression (14), can be written as:

(17)    

Rearranging to isolate the impact of monetary policy gives

(18)    

The equivalent expression for output variability, drawing on equations (10) and (11), is

(19)    

Rearranging to isolate the impact of monetary policy gives

(20)    

The impact of monetary policy on the output and inflation variability trade-off is interpreted as the impact on the contemporaneous pairing of the realised values in each quarter for equations (18) and (20). These are shown in Figure 5 which comprises three panels: the top panel is the period 1983 to 1989, immediately prior to the introduction of the Reserve Bank of New Zealand Act 1989; the middle panel is the period 1990 to 1995, the initial years of formal inflation targeting; the bottom panel is the period 1996 to 2001, the last six years of inflation targeting covered by the model estimation.

The realised values for expressions (18) and (20) can be either positive or negative. A positive realised value implies that monetary policy accentuated variability. A negative value implies that monetary policy reduced variability. To maximise a loss function of the type shown by expression (16), in the presence of other shocks to output and inflation, ideally observations for expressions (18) and (20) should all lie in the south-west quadrant of each panel.

In the period 1983 to 1989, prior to the advent of formal inflation targeting, the realised observations for equations (17) and (19) are scattered over all quadrants. There are more observations above the horizontal zero axis meaning that monetary policy was predominantly adding to inflation variability. The observations for output variability lay predominantly close to or left of the vertical zero axis. This indicates that monetary policy was predominantly reducing output variability.

Figure 5 – Monetary policy contribution to output and inflation variability before and during inflation targeting

^{Note: Each scatter point represents the contemporaneous realisations of monetary policy’s contribution to output and inflation variability. The scatter points in the third chart that are represented by square markers show monetary policy’s contribution during the MCI period.}

A different pattern of realisations emerges for the period after formal inflation targeting was introduced. Moreover, the pattern for the first six years, shown in the middle panel of Figure 5 as 1990-1995, is different to the pattern for the second six years, which is shown in the bottom panel as 1996-2001.

In the first six years of inflation targeting (shown in the middle panel of Figure 5), many of the realisations lie in the south-west quadrant where policy reduces both inflation and output variability. All of the inflation variability realisations lie close to or below the horizontal axis implying that monetary policy was predominantly reducing inflation variability. The output realisations are scattered on both sides of the horizontal axis, but predominantly in the south-west quadrant. There are however some output variance realisations that are higher than the output variance realisations arising from the impact of monetary policy in the period prior to inflation targeting. The marked reduction in inflation variability in the first six years of inflation targeting is consistent with the idea of a change in the Reserve Bank’s preference function with a higher preference given to the reduction of inflation variability compared to the earlier period.[18]

The fact that inflation variability was reduced without any significant adverse increases in output variability is an outcome consistent with enhanced monetary policy credibility. According to this argument, once lower inflation expectations are achieved, shocks that move inflation away from target can be brought back to target with less cost in terms of output variability. This outcome is also consistent with the findings of Fischer and Orr (1994) and Hutchison and Walsh (1998) who provide empirical evidence of improved monetary policy credibility during the early years of inflation targeting in New Zealand. Improved policy credibility also provides the opportunity for a central bank to shift its emphasis to flexible inflation targeting, an interpretation made by Svensson (1997).

Accordingly, we might expect to find a similar or even improved set of realisations during the second six years of inflation targeting. These realisations are shown in the bottom panel of Figure 5. The impact of monetary policy during the second six years of inflation targeting is not as favourable as the realisations for the first six years. Most of the observations lie close to and right of the vertical axis, implying monetary policy predominantly increased output variability in this period. The realisations for the impact on inflation variability are closely bunched around zero, and some realisations added to inflation variability to a greater extent than observed in the previous six-year period.

There are several possible explanations for these less favourable outcomes in the period 1996 to 2001. The Reserve Bank was operating monetary policy using the Monetary Conditions Index during part of this period. At no stage during the period when the MCI was operating did monetary policy reduce output variability and for all but two quarters it increased output volatility. Between 1996 and 2001 New Zealand experienced two large climate shocks whose effects on output may have been difficult to interpret. In addition, there was less scope for monetary policy to reduce variability as output and inflation variability from other sources were lower in this latter period.

The reduction of inflation variability over time has coincided with a reduction in the stochastic trend rate of growth of domestic prices. Similarly, trend output growth increased during the 1990s. The SVAR model used to evaluate the consequences of inflation targeting is a model of the growth cycle and does not model trend output and trend inflation. It may be the case that the observed fall in inflation and output variability is in part attributable to changes in trend growth for these variables. In this situation, estimates of monetary policy’s contribution to reducing output and inflation variability may be biased downward.