Abstract
As inflation targeting gains popularity policy makers, monetary authorities seek to design a measure of inflation that would be a good indicator of fundamental demand-driven price movements, i.e. the underlying or core rate of inflation. It is widely acknowledged that the Consumer Price Index (which is the simple weighted average of price changes of the set of goods and services comprising the consumers' expenditure basket) is a rather deficient indicator of the 'trend' inflation as it is highly volatile, seasonal and contains a lot of noise. The ideal measure of core inflation should account for the long-term trend movements in prices that reflect the state of demand in the economy and discard various one-off shocks coming from supply side. The paper presents 4 alternative methods of calculating the core inflation most commonly found in the literature: trimmed mean, sample mean percentile, standard deviation trimmed mean and exclusion mean. Using Polish price data from the period 1995:1-1998:7, each measure is calculated at monthly, quarterly and annual frequencies and compared to the 24-month centered moving average .of the CPI which is assumed to be the benchmark core inflation. Root mean square error (RMSE) and mean absolute deviation (MAD) of the candidate measure and the benchmark were chosen to be the criteria for choosing the optimal definition - both within each of the 4 groups and across them. Rather surprisingly, crude methods based on exclusion yielded the best results. Volatility-based exclusion proved most efficient for monthly and quarterly series, whereas excluding broad aggregates (food and energy) turned out optimal for annual series. The paper concludes with highlighting the caveats and fragility of the results as well as stressing the necessity of further research.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have