Ignoring items with large price changes may enhance the informational content of a price index. As an application of the metrically trimmed mean (Kim, 1992) we suggest to discard the individual price changes that deviate the most from the median. Focusing on outliers increases the efficiency compared to always trimming equally in both tails and the implied bias problem seems small. The distribution of price changes is often skewed strongly to the right or to the left in a specific month but is much closer to symmetric for a longer period as a whole. This is also the case with Danish data analyzed in this paper. The suggested metrically trimmed mean gives a measure of expected inflation, which may help representing inflation in economic analyses.