A number of measures have been developed to measure underlying inflation. The CPI excluding food and energy prices (core CPI) removes price changes of the same items from the CPI each month—namely, food and energy prices—because they are typically the most volatile. The median CPI and the trimmed-mean CPI use a different approach. These measures exclude the smallest and largest price changes during the month, so the items excluded from the CPI change from month to month. The median CPI excludes all price changes except for the one in the center of the distribution of price changes, where the price changes are ranked from lowest to highest (or most negative to most positive). The 16 percent trimmed-mean CPI excludes price changes in specified upper and lower tails of the distribution.
According to research from the Cleveland Fed, the median CPI provides a better signal of the underlying inflation trend than either the all-items CPI or the CPI excluding food and energy. The median CPI is even better at forecasting PCE inflation in the near and longer term than the core PCE price index.
How Median CPI and 16 Percent Trimmed-Mean CPI Are Calculated
To calculate the median CPI, the Federal Reserve Bank of Cleveland looks at the prices of the goods and services published by the Bureau of Labor Statistics (BLS). But instead of calculating an inflation rate that is a weighted average of all of the items in the CPI, as the BLS does, the Cleveland Fed ranks the inflation rates of the components of the CPI and picks the one in the middle of the distribution—that is, the item whose expenditure weight is in the 50th percentile of the price change distribution. The Cleveland Fed also calculates the 16 percent trimmed-mean CPI by taking a weighted average across the component inflation rates after excluding, or trimming, items whose expenditure weights fall in the top 8 percent and bottom 8 percent of the price change distribution.
The method used to calculate the median CPI and the 16 percent trimmed-mean CPI was revised in September 2007 to correct for the potential distortion that the owners’ equivalent rent (OER) component could cause to the measures. For more detail, see this document.