Does the RBA Already Follow an Implicit, Forward-Looking Nominal GDP Targeting Rule?

My RBA Governor NGDP Targeting Beauty Contest

In considering the merits of a nominal income targeting rule for Australia, it is worth considering whether such a rule is already a good description of the RBA’s monetary policy reaction function. The RBA has been shown to follow a forward-looking Taylor rule similar to those estimated for the US Fed and other central banks in Clarida, Gali and Gertler (1998). De Brouwer and Gilbert (2005) estimate forward-looking rules of this type for the RBA and find a weight on inflation of around 2.3 and on the output gap of between 0.7 and 0.98. As the great Barry Hughes (1997) once suggested at a RBA conference, the RBA ‘behaved throughout the period since 1984 as sadistically as any Taylor rule would have demanded.’

A nominal income targeting rule can be thought of as simply a re-weighted Taylor rule, with a single weight on the combination of inflation (as measured by the GDP deflator) and output deviations. While the RBA traditionally thinks of itself as implementing something like a Taylor rule, that does not mean its reaction function is inconsistent with nominal income targeting.

A nominal income targeting rule can be written as (1):

where i is the target official cash rate, α is a constant term, β is the elasticity of the official cash rate target to deviations in nominal GDP from its expected value at time t+n and ρ is an interest rate smoothing parameter designed to capture hedging behavior or inertia on the part of monetary policy.

I use the NGDP gap estimated in this previous post, updated for the March quarter 2020 national accounts release. Obviously, this is only one of many ways we could potentially measure deviations in actual NGDP from expectations and is far from unimpeachable. Implicitly, we are assuming that people form their expectations for nominal income based on this measure. Feel free to substitute your own.

I estimate (1) using generalised method of moments (GMM) with a two-quarter forecast horizon on the nominal GDP gap. The instruments used for estimation are lags of NGDP gap (-1), the official cash rate (-2), the log first-difference of the AUD-USD exchange rate (-1, -2) and the US effective Fed funds rate (-1, -2).

I consider four sample periods, covering the terms of Governor Macfarlane; Governor Stevens; Macfarlane and Stevens combined; and Macfarlane, Stevens and Lowe combined. Lowe has been in office for only four years and left the cash rate target unchanged during the first three, making it difficult to estimate a separate reaction function for his time in office using quarterly data. The estimated parameters for each sample period are shown in Table 1 below.

Looking at the estimates by Governor, we find the long-run response to the NGDP gap under Governor Macfarlane is 1.37, although this is not statistically significant. However, the model has the lowest SE under Governor Macfarlane, suggesting his term otherwise has slightly greater adherence to a NGDP gap targeting norm. Stevens has a larger response to the NGDP gap of 1.75 and this is statistically significant. Combining Macfarlane and Stevens’ terms (Mac-Stevens), the response is very similar at 1.69.

Adding Governor Lowe’s term to his two predecessors sees a rise in the interest rate smoothing parameter as Lowe puts the official cash rate target into a nearly three-year coma (De Brouwer and Gilbert (2005) observed that Australian monetary policy is characterised by ‘deep stasis’). This rather mechanically raises the long-run response to the NGDP gap, but this is more a reflection of the increased inertia of monetary policy. The J-statistic and associated p-value accepts the null hypothesis that the over-identifying restrictions are satisfied for each sample period.

These reduced-form estimates can only be viewed as representative of the RBA’s preferences under each Governor(s) if the structure of the economy is unchanged or the instruments adequately account for the endogeneity of any structural change. We can assume the terms of individual RBA Governors are exogenous to economic conditions. Otherwise, the varying estimates can be interpreted as a change in the response of monetary policy to a change in the structure of the economy. It could be that monetary policy has had to work harder to stabilise nominal GDP over time and this is reflected in the cash rate’s increased response to the NGDP gap. But the coefficient on the NGDP gap appears remarkably stable under Macfarlane and Stevens.

It could also be the case that the forward-looking NGDP targeting rule is simply recovering an embedded forward-looking Taylor rule. This is a familiar problem in the literature on estimated monetary policy rules. In any event, a forward-looking NGDP targeting rule is still a good empirical description of Australian monetary policy for the period since the inflation target was formalised in 1996, suggesting that a shift to nominal GDP targeting is not a radical departure from past practice and may offer operational advantages over a Taylor rule, which depends far more heavily on unobserved variables relative to a nominal income targeting rule.