RDP 2006-11: Component-smoothed Inflation: Estimating the Persistent Component of Inflation in Real Time 6. Conclusion

The component-smoothed inflation measure of underlying inflation developed in this paper approaches the problem of removing noise from headline inflation in quite a different way than existing measures of underlying inflation. Instead of reweighting the component price series, we smooth them in proportion to the amount of noise they contain to preferentially remove temporary shocks; volatile series such as fruit and automotive fuel are smoothed heavily while stable series such as rents are smoothed very little, with other CPI items lying on a continuum between these extremes. This leads to a measure that has a number of desirable properties for an underlying measure of inflation: it is unbiased with respect to headline inflation in the long run, it can be calculated in real time and it is much smoother and less noisy than headline inflation. Additionally, as was discussed with respect to the US experience, where the measure does lag headline because of persistent shocks to historically volatile series, this lag may be appropriate.

An additional feature of the component-smoothed inflation measure is that it requires a relatively small number of parameters. Thus, while we have specified certain values for these parameters in this paper, these can easily be respecified by the user, depending on the desired smoothness of the component-smoothed inflation measure. Notwithstanding this, the general framework of component smoothing based on volatility advanced in this paper allows for more advanced smoothing mechanisms to be employed; while we suggest that the exponential smoothing mechanism provides a good trade-off between simplicity and rigour, the use of the Kalman filter to smooth the component series is an easy modification for more technically inclined users.

In addition to having desirable general properties, the results for Australia and the US provide estimates of underlying inflation that offer potentially useful perspectives on inflationary pressures. The component-smoothed inflation measure tracks the medium-term trends in headline inflation that are relevant for monetary policy formulation, while removing – in real time – much of the noise present in headline inflation.

Notwithstanding these positive attributes, as demonstrated in the tables of results comparing the various underlying inflation measures, while exclusion measures are usually dominated, none of the component-smoothed inflation and statistical measures are better than each other in all respects. In these circumstances each can offer a different and valuable perspective on underlying inflationary pressures. Thus, by approaching the problem of measuring underlying inflation in a different way the component-smoothed measure may provide a useful supplement to existing measures. An evaluation of the balance between the costs and benefits of this approach remains an open question.