What is trimmed mean inflation?

Trimmed mean inflation is a statistical technique that removes the items with the largest and smallest price changes from the CPI or PCE basket each month, then calculates the weighted average of the remaining items. The most widely followed version is the Dallas Fed's trimmed mean PCE inflation rate, which trims the top and bottom 24% (by expenditure weight) of monthly price changes, leaving the middle 52% to determine the underlying inflation trend.

The logic behind trimmed mean is straightforward. In any given month, a handful of CPI components may experience extreme price swings: airline fares might spike 15% due to jet fuel costs, or used car prices might plunge 4% due to a fleet liquidation. These outliers can heavily distort headline and even core inflation readings. By symmetrically removing the tails, the trimmed mean isolates the central tendency of price changes across the entire basket.

Research at the Dallas Fed shows that trimmed mean PCE has been a more accurate predictor of future inflation than both headline and core measures. It was particularly valuable during 2021-2022, when a few categories (used cars, energy, food) distorted headline inflation dramatically. The trimmed mean signaled persistent underlying inflation before core measures fully reflected it, and it subsequently signaled the disinflation trend earlier as well.

For practitioners, the trimmed mean offers a cleaner signal for monetary policy assessment. If the trimmed mean is running well above 2% annualized, it suggests that broad-based price pressures exist regardless of what headline or core readings show. If the trimmed mean is falling even while a few sticky categories hold core CPI elevated, it suggests the disinflation process is progressing. The Cleveland Fed produces a similar measure, the median CPI, which takes the single median price change across all categories, offering an even more stripped-down view of underlying inflation.