RDP 2006-11: Component-smoothed Inflation: Estimating the Persistent Component of Inflation in Real Time 5. Caveats

As with all underlying inflation measures, our component-smoothed inflation measure has strengths and weaknesses. We have highlighted some of its strengths in preceding sections. This section highlights areas where the measure needs to be interpreted with caution.

The measure makes assumptions about the behaviour of individual price series based on their historical behaviour; if that behaviour changes, those assumptions will be inaccurate. This problem is most relevant when evaluating series that are judged to be volatile.

Movements in series which have been persistent in the past are reflected in the measure quickly. As such, temporary shocks to a persistent series will be associated with marginally more volatility in the measure but are unlikely to mislead about the underlying rate of inflation. Notwithstanding this, in such circumstances, a measure like the trimmed mean, which trims away large movements regardless of the series in which they occur, could have lower volatility.

On the other hand, if there is a permanent shock to a volatile series, the component-smoothed measure will be slow to incorporate this. We have discussed the US experience of energy price fluctuations where this is argued to be beneficial. Nonetheless, shocks to other prices may not affect the rest of the economy in the same way as energy prices do and, as such, the component-smoothed inflation measure may be slower to reflect these inflationary pressures than the CPI itself or may reflect shocks whose influence on the economy has already dissipated.

A related issue is whether an optimal measure of current inflation should be affected by movements from previous periods. An argument can be made that, in essence, bygones should be bygones and large inflationary shocks in previous periods should not continue to directly influence a measure of underlying inflation in and of themselves. Our measure, because it is calculated on the level and maintains unbiasedness with respect to the CPI, does not have this property. Rather, the measure will display higher inflation than usual in the immediate aftermath of a large sustained positive price shock to a historically volatile series. Another way of putting this issue is to ask whether an optimal measure of current inflation needs to be unbiased over time. The resolution of that issue is left for other work.