Trimmed Means of the CPI as an Indicator of Core Inflation for Costa Rica
MetadataShow full item record
This study presents an evaluation of trimmed-mean inflation series calculated from Costa Rican data following the methodology of Bryan, Cecchetti and Wiggins (1997) and of Roger (1997), which entails the calculations of trimmed means centered in an estimator of the population mean percentile. It was found that the historical distribution of price changes in Costa Rica is highly leptokurtic and right-skewed, the mean percentile being estimated at percentile 60th. Trimmed-mean series were calculated by centering on the percentiles 50th to 70th and by using trimming percentages from 0% to 49%. In order to choose a trimmed-mean indicator for core inflation, the series obtained were evaluated through unbiasedness tests, forecasting ability tests and indicators of adequacy of fit to a measure of trend inflation. Most unbiased series were found to be centered around the estimated mean percentile. The series resulting from centering on the mean percentile (60th) and trimming 30% of the weight on the left of the distribution and 10% on its right presents the best fit to the trend among the group of 24 unbiased series that showed the highest forecasting ability. This trimmed-mean series fits better to the trend inflation than an exclusion measure currently in use (the ISI, for its Spanish acronym), and its variability is lower.
- Documentos de Trabajo 
This work is licensed under a Creative Commons Reconocimiento-NoComercial 4.0.