RDP 2010-03: Modelling Inflation in Australia 7. Conclusions

Given the goal of central banks to maintain low and stable inflation, researchers devote significant effort to modelling inflation. This paper adds to the existing literature by exploring some of the single-equation models that are used at the RBA.

A key finding of our paper is the strong performance of the standard Phillips curve and mark-up models when compared to the New-Keynesian Phillips curve over the past two decades. In particular, we find that the issue of weak instruments in the NKPC results in coefficients that are significant and correctly signed only for the inflation expectations and lagged inflation variables: the apparent insignificance of the output gap and marginal cost terms is problematic (though this is commonly also an issue in US research). Furthermore, the fit of the NKPC is clearly inferior to that of the standard Phillips curve and mark-up models. These issues with the NKPC can be somewhat alleviated by the use of a direct measure of inflation expectations, which avoids the need to instrument for this variable, although it remains the case that the more traditional models out-perform this OLS-based version of the NKPC.

One issue with the standard Phillips curve and mark-up models is that unrestricted estimates imply a medium-term trade-off between output and inflation (and fit less well when a restriction is imposed to prevent this). While this is unlikely to be a major issue when inflation is at relatively low levels, this aspect could be more problematic in the case of more extreme events. For this reason, any suite of models should include a role for other types of more theoretically based models that naturally incorporate such restrictions.

Regardless of the model chosen, it is notable that the standard error of single-equation models for inflation has fallen in the low-inflation environment prevailing over the past 15 or so years, consistent with the stability of inflation and inflation expectations over that time.