Why does the Fed watch median and trimmed-mean CPI?

The short answer

Headline CPI moves with food and energy prices, which are highly volatile. Core CPI removes these mechanically, regardless of whether they are actually behaving unusually in any given month. This works most of the time but creates blind spots: in a month when energy is well-behaved but used cars surge by 20%, core CPI captures the surge as if it were broad inflation.

Median and trimmed-mean approaches solve this differently. Instead of removing predetermined categories, they remove categories that are behaving unusually in that specific month — high or low. The result is a measure that captures the central tendency of price changes, not the average distorted by outliers.

The Cleveland Fed has published Median CPI since 1991. The Dallas Fed introduced its Trimmed Mean PCE in 2005. Both have become standard reference measures in FOMC discussions, particularly during periods when individual categories produce extreme moves.

→ New to underlying inflation measures? Financial education hub

What the data shows

Key figures (Cleveland Fed, Dallas Fed, FRED, 2010-2024):

• Cleveland Fed Median CPI peaked at 7.2% in October 2022, vs core CPI peak of 6.6% in September 2022
• Dallas Fed 12-month Trimmed Mean PCE peaked at 4.7% in September 2022, vs core PCE peak of 5.4% in February 2022
• Trimmed Mean PCE typically removes about 24% of components by weight each month (12% from top, 12% from bottom)
• Median CPI captures the 50th percentile of price-change distribution across approximately 45 expenditure categories
• The 2021-2022 episode saw median CPI lead core CPI in signaling broad-based inflation acceleration by approximately one quarter

The 2021-2022 episode demonstrated the value of these measures: when used cars and shelter were producing extreme moves, median CPI showed the broad-based nature of inflation earlier than headline or core measures.

→ Dataset: US Core PCE

Why it happens — the macro mechanism

Three reasons explain why central banks complement core CPI with these measures.

The fixed exclusion problem. Core CPI removes food and energy regardless of whether they are the source of unusual inflation. In 2021, the unusual movers were used cars and apparel; in 2022, shelter and services. Core CPI captured these as part of underlying inflation, even though they were largely idiosyncratic supply phenomena. Median and trimmed-mean approaches respond month by month to whichever components are extreme. See core vs headline inflation.

Cross-sectional distribution. In any given month, prices of approximately 200 narrow CPI components form a distribution. The shape of this distribution carries information: a wide distribution suggests idiosyncratic shocks; a narrow distribution suggests broad-based pressure. Median CPI captures only the central tendency, ignoring tails. The width of the distribution itself becomes a separately useful indicator. See how CPI is calculated.

Symmetric trimming. Trimmed-mean approaches remove both unusually high and unusually low movers. This corrects for a known issue with median statistics: they discard information from non-outlier components. The Dallas Fed’s choice of asymmetric trimming (different percentages from top vs bottom) was empirically calibrated to maximize the predictive power for future PCE inflation. See inflation expectations.

Median and trimmed-mean inflation are not better than headline — they answer a different question, and that difference is why central banks watch both.

→ Framework: Inflation regimes pillar

What it means for different economic actors

Savers who use headline CPI to gauge purchasing power may want to cross-check with median CPI to understand whether a particular CPI print reflects broad-based pressure or category-specific noise. The two often diverge meaningfully in volatile periods.

Investors in fixed income use these measures to anticipate Fed policy moves. The Cleveland Fed Median CPI has shown empirical value as a leading indicator of FOMC reaction function changes, particularly when it diverges from core CPI. Bond markets often re-price expectations following monthly median CPI releases. See breakeven inflation rates.

Policymakers at the Fed reference these measures explicitly in FOMC minutes. The 2021-2022 transcripts show several participants citing median and trimmed-mean measures as evidence that inflation breadth was widening before headline measures fully reflected it. Powell’s 2024 speeches have referenced trimmed-mean PCE specifically. See Fed 2% target.

A common misreading is treating median or trimmed-mean as the “true” inflation. They are alternative views, each useful for specific questions. Headline remains the relevant measure for purchasing power; core and median remain relevant for underlying trend assessment.

Practical observation

Go deeper

Related questions

• How is CPI actually calculated?
• Core vs headline inflation
• PCE vs CPI inflation difference
• Why are inflation expectations self-fulfilling?

Frequently asked questions

Is median CPI better than core CPI?

Neither is universally better — they serve different analytical purposes. Core CPI removes a fixed set of categories (food and energy) every month, which is simple and predictable but can miss idiosyncratic moves in other categories. Median CPI dynamically removes whatever is extreme, which adapts to changing inflation drivers but is harder to interpret. The Cleveland Fed publishes both, and most economists examine them together.

Why does the Dallas Fed use trimmed mean for PCE rather than CPI?

The Dallas Fed’s choice reflected the Fed’s official focus on PCE as the policy target. The methodology is similar to trimmed mean CPI but applied to PCE expenditure categories with PCE weights. The asymmetric trimming (different percentiles top vs bottom) was empirically optimized to forecast future headline PCE inflation, providing a real-time gauge of underlying pressure relevant to the FOMC’s mandate.

Do these measures replace traditional core inflation?

No, they complement it. Core inflation remains the most widely cited underlying inflation measure in financial media and policy communications because of its longer history and simpler interpretation. Median and trimmed-mean measures are typically referenced when headline and core diverge meaningfully, or when individual categories produce extreme moves that complicate the narrative. Most FOMC participants reference multiple measures in their economic projections.

Last updated — 1 May 2026