RDP 2009-06: Inflation Volatility and Forecast Accuracy 5. Conclusion

This paper compares the statistical properties of inflation in a sample of IT and non-IT countries using an unobserved components stochastic volatility model (UC-SV) proposed by Stock and Watson (2007). This approach decomposes inflation into permanent and transitory components and allows the variability of both components to change over time. We find that IT countries (Australia, Canada, New Zealand, Sweden and the United Kingdom) on average experienced a more pronounced fall in the volatility of the permanent component of inflation than non-IT countries (Austria, France, Germany, Japan and the United States), to levels that are now broadly comparable.

We also propose a modification of the original UC-SV model in order to allow for more plausible implied properties of inflation which are particularly desirable when forecasting inflation. We find that inflation became easier to forecast in both IT and non-IT countries since the early 1990s, but forecast errors remained somewhat smaller in the latter group. Across the countries in our sample, forecasts from the modified UC-SV model are generally superior to those based on the assumption that the eight-quarter-ahead inflation rate is the average rate of inflation over the previous four quarters.