Every Major U.S. Inflation Measure, Explained

The basics

There isn't one 'inflation number.' Different measures look at prices differently and often disagree short-term.

When you read a news headline that says "inflation hit 3.2%," the natural assumption is that there's a single official number that everyone uses. There isn't. The U.S. government publishes more than a dozen different inflation measures, each constructed slightly differently, each emphasizing different aspects of the price-change story. They mostly agree over long time horizons, but on any given month they can diverge by half a percentage point or more.

The most familiar one is CPI - the Consumer Price Index - published by the Bureau of Labor Statistics. That's what most news headlines cite. The most important one for monetary policy is PCE - Personal Consumption Expenditures price index - published by the Bureau of Economic Analysis. That's the measure the Federal Reserve targets at 2%. Each has a "core" version that strips out food and energy to filter short-term noise.

A few specialty measures try to do that filtering better. The Cleveland Fed publishes the Median CPI, which takes the middle component of the basket each month - removing both the most extreme price increases and the most extreme decreases. The Atlanta Fed publishes Sticky CPI, which includes only items whose prices change infrequently (rent, education, medical services) and excludes things that change every week (gasoline, vegetables). The Dallas Fed publishes Trimmed Mean PCE, which drops the most extreme price movements in both directions before averaging. Each is a different way to find the underlying inflation signal beneath the monthly noise.

And then there's the forward-looking family. Breakeven inflation rates from the bond market tell you what investors are willing to bet on. Surveys from the University of Michigan and the New York Fed tell you what households expect. The Survey of Professional Forecasters tells you what economists are predicting. These don't measure realized inflation - they measure expected inflation, which matters because inflation expectations themselves influence wage demands, business pricing, and ultimately realized inflation in a self-fulfilling loop.

Together, the full set paints a much richer picture than any one number alone. When someone tells you "inflation is at X%," the right follow-up is "which measure?" and "compared to what?" The answer can substantially change the policy implication.

Going deeper

Three families: realized indices (CPI, PCE), underlying measures (Trimmed, Sticky, Median), expectations (breakevens, surveys).

U.S. inflation measures fall into three families. Each family answers a different question, and the right policy or analytical interpretation depends on knowing which family you're looking at.

Realized indices - the actual price changes that have happened. CPI Headline (FRED CPIAUCSL) and CPI Core (CPILFESL) are the BLS's flagship measures, with the Headline tracking all items and Core stripping food and energy. PCE Headline (PCEPI) and PCE Core (PCEPILFE) are the BEA's analogs, with Core PCE being the Fed's official 2% target measure. CPI tends to print 0.3–0.5 pp higher than PCE on a year-over-year basis because of formula and scope differences (see the CPI vs Core CPI and PCE vs Core PCE explainers for detail). C-CPI-U (Chained CPI) uses a Törnqvist superlative formula and tends to print closer to PCE.

Underlying / trend measures - alternative central tendencies built from the cross-section of component price changes each month. The point is to filter out short-term shocks (a single component spiking or crashing) while capturing the underlying trend. Trimmed Mean PCE (FRED PCETRIM12M656N, Dallas Fed) drops a fixed 24% of expenditure weight from the lower tail and 31% from the upper tail of the monthly component distribution (about 55% total, asymmetric because the price-change distribution is negatively skewed) and averages what's left. Sticky-Price CPI less food and energy (CORESTICKM159SFRBATL, Atlanta Fed) includes only components whose prices change less frequently than every 4.3 months. Median CPI (MEDCPIM158SFRBCLE, Cleveland Fed) takes the 50th-percentile component change weighted by expenditure share. 16% Trimmed-Mean CPI (TRMMEANCPIM158SFRBCLE, Cleveland Fed) drops the top 8% and bottom 8% of components.

Expectations - what people think inflation will be. Market-based: 5-year breakeven inflation (T5YIE), 10-year breakeven (T10YIE), 5-year-5-year forward breakeven (T5YIFR - the Fed's favorite for isolating long-run expectations). Survey-based: University of Michigan Survey of Consumers (MICH for 1-year) is the longest-running consumer-expectations survey; New York Fed Survey of Consumer Expectations (SCE, post-2013) provides 1-year, 3-year, 5-year horizons with demographic breakdowns; Survey of Professional Forecasters (SPF, Philadelphia Fed, quarterly) provides 1-year, 5-year, 10-year forecasts from economists. Model-based: Cleveland Fed Expected Inflation series derives implied inflation from inflation swaps and other inputs; ACM-Inflation (NY Fed) decomposes breakevens into expected inflation and inflation risk premium.

When the families align, the inflation signal is robust. When they diverge, each tells you something different. Market-based expectations can move on Fed credibility shifts even when realized inflation hasn't changed. Survey-based expectations move on household experience (gas prices, grocery prices). Underlying measures filter the noise from realized inflation but can lag turning points. Realized indices are what's actually happening to prices today.

A quick worked example. Suppose Core CPI year-over-year is 3.5%, Core PCE is 3.0%, Trimmed Mean PCE is 3.2%, Sticky CPI is 4.0%, Median CPI is 3.6%, 5Y breakeven is 2.4%, 10Y breakeven is 2.3%, T5YIFR is 2.2%, Michigan 1Y is 3.2%. The story: realized inflation is around 3% with shelter-related stickiness elevating CPI-based measures; market-based long-run expectations are well-anchored near 2.2–2.4%; consumer expectations are slightly above realized. Policy interpretation: the inflation overshoot is moderating but service-sector stickiness persists; market credibility is intact.

Advanced detail

CPIAUCSL / CPILFESL / PCEPI / PCEPILFE / PCETRIM12M656N / CORESTICKM159SFRBATL / MEDCPIM158SFRBCLE / T5YIE / T10YIE / T5YIFR / MICH.

Comprehensive FRED reference list for U.S. inflation analysis. Realized indices: CPIAUCSL (CPI-U headline SA), CPIAUCNS (CPI-U headline NSA, for TIPS indexation), CPILFESL (CPI-U core SA), CPILFENS (core NSA). CUUR0000SA0 (CPI-U headline NSA, official BLS), CWUR0000SA0 (CPI-W headline NSA, used for SS COLA), SUUR0000SA0 (Chained CPI-U, all items, NSA). PCEPI (PCE headline SA), PCEPILFE (PCE core SA). Goods-vs-services: DGDSRG3M086SBEA (goods PCE prices), DSERRG3M086SBEA (services PCE prices); core goods and core services decompositions come from BEA's underlying detail tables.

Underlying-inflation measures: PCETRIM12M656N (Dallas Fed Trimmed Mean PCE, 12-month change), MEDCPIM158SFRBCLE (Cleveland Fed Median CPI, 1-month annualized), CORESTICKM159SFRBATL (Atlanta Fed Sticky-Price CPI less food and energy, y/y % change), TRMMEANCPIM158SFRBCLE (Cleveland Fed 16% Trimmed-Mean CPI), STICKCPIXSHLTRM159SFRBATL (Atlanta Fed Sticky-Price CPI excluding shelter, y/y % change).

Expectations measures: T5YIE (5Y breakeven), T10YIE (10Y breakeven), T5YIFR (5Y5Y forward breakeven), DFII5/7/10/20/30 (constant-maturity TIPS yields). MICH (Michigan 1-year inflation expectation, monthly); Michigan 5-10 year expectations are published by the Surveys of Consumers directly, not on FRED. SPF data via Philadelphia Fed's quarterly release (separate from FRED for some series). NY Fed SCE data via NY Fed's published time series (not on FRED directly).

CPI vs PCE structural gap analysis: McCully-Moyer-Stewart (2007, BEA Survey of Current Business) decomposition attributes the gap to formula (Fisher chain vs modified Laspeyres), weights (quarterly NIPA for PCE vs CEX-based CPI weights, updated annually since 2023 and biennially before that), and scope (PCE includes employer-paid medical, financial services, certain imputations CPI excludes); the components partially offset in some periods rather than simply adding up. The gap shifts by year - during medical-care-cost-spike periods, the medical-scope gap widens; during high-energy-price periods, the formula gap shrinks (less substitution opportunity).

Underlying-measure construction details: Trimmed Mean PCE - Dallas Fed drops the tails of the monthly PCE component price-change distribution, weighted by expenditure share. The trim is fixed and asymmetric - 24% of expenditure weight from the lower tail, 31% from the upper tail, about 55% total - with proportions chosen in Dolmas (2005) and updated in 2009 to match the negatively skewed price-change distribution. 12-month change is the standard reporting metric. Sticky CPI - Atlanta Fed defines "sticky" as items whose prices change less frequently than every 4.3 months (the median frequency in the CPI sample). The cutoff was set in Bryan-Meyer (2010); robustness has been tested with alternative cutoffs. Median CPI - Cleveland Fed orders the CPI component price changes each month by magnitude (weighted by expenditure share) and selects the 50th-percentile change. 1-month annualized is the most-noted figure; 12-month change is also published.

Market-vs-survey-vs-model expectations triangulation: Fed staff inflation outlook (Tealbook materials, declassified after 5 years) uses all three families. When all align, the signal is treated as robust. When they diverge, each is investigated for its specific drivers. Market measures (breakevens) move on liquidity, Fed credibility, supply-shock perceptions. Survey measures move on household price experience (gas, groceries - disproportionate weight). Model-based measures (Cleveland Fed, ACM-Inflation) blend market and survey inputs with model structure.

The operational policy implication of triangulation: the 2022 episode saw the market-based 5Y5Y forward breakeven briefly break above 2.5% in April 2022 while survey-based long-run measures (Michigan 5-10Y, NY Fed SCE 3Y) split by horizon: Michigan 5-10Y held around 2.9–3.1%, whereas the NY Fed SCE 3-year measure (a medium-term gauge rather than long-run) rose well above 3%, peaking near 4.2% in late 2021 before receding. Elevated readings across both the market and survey families raised the perceived threat. The successful 2023–24 disinflation pulled both families back near 2.0–2.5%.

Supercore PCE (core services ex-housing) - analytical decomposition rather than a separate published index. Constructed by subtracting housing services and energy services from core services. Roughly 51% of the core PCE basket. The wage-link makes supercore the policy-relevant measure for assessing whether the disinflation has reached the underlying-trend level - when wage growth is above the rate consistent with 2% inflation given productivity, supercore inflation tends to stay elevated.

OER lag in CPI shelter (12–18 months behind market rents) was the central operational issue for the 2022–24 disinflation interpretation. Market-rent indices (Zillow ZORI, Apartment List, CoreLogic SFRI, BLS New Tenant Rent Index) decelerated rapidly through 2023, but CPI Shelter remained elevated through most of 2024 as the older high-rent observations in the BLS sample rolled out. PCE Housing uses a different methodology with somewhat different timing.

Research caveat: comparing inflation measures across time requires checking methodology revisions. BLS made significant changes to medical-care methodology in 2023 (new health-insurance treatment); PCE annual revisions can shift historical data by 0.2–0.5 pp.

Expert notes

Fed staff cross-checks all three families; supercore PCE (services ex-housing) for wage-driven persistence.

Operational policy view: Chair Powell's November 2022 Brookings speech ("Inflation and the Labor Market") explicitly highlighted three categories within core PCE: core goods (sensitive to supply shocks, tariffs, FX, semiconductor cycle), housing services (driven by OER lag), and non-housing core services ("supercore" - wage-driven). The supercore category was identified as the most policy-relevant because of its tight wage-growth correlation and historical disinflation stickiness. The speech is the canonical reference for how the Fed has framed the inflation decomposition through 2022–25.

Realized-measure caveats. (1) OER lag in CPI shelter - 12–18 months - was the 2022–24 disinflation story's gating factor. The pre-2023 rapid market-rent surge produced an OER spike that didn't fully unwind until late 2024 / early 2025. (2) Health-insurance methodology change - BLS implemented a methodology change in October 2023 that produced a step change in measured health-insurance prices. (3) BEA annual revisions can shift PCE history materially - the 2023 revision marked up 2022 Core PCE by approximately 0.4 pp. (4) Goods-services rotation post-COVID - the COVID-era goods-spending spike followed by services-spending normalization produced unusual component dynamics that distorted historical comparison.

Expectations-measure caveats. (1) TIPS liquidity premium depresses breakevens by roughly 10–40 bps in normal conditions, larger in stress; the deflation floor works the other way, pushing observed breakevens up in low-inflation regimes. (2) Survey measures show systematic biases: Michigan 1-year consistently runs above realized inflation (consumer perception is anchored to recent high-frequency price experience - gasoline, groceries); SPF tends to under-forecast during turning points. (3) Demographic dispersion in NY Fed SCE shows lower-income, lower-education respondents systematically expect higher inflation than higher-income, higher-education respondents.

Decomposition tools. D'Amico-Kim-Wei (2018) is the standard decomposition of breakevens into expected inflation, inflation risk premium, and liquidity premium, with regularly updated estimates published by the Federal Reserve Board. The Abrahams-Adrian-Crump-Moench-Yu (2016) NY Fed staff report provides an alternative decomposition. Cleveland Fed publishes a model-based expected inflation series that uses the Cleveland Fed's affine model fit to inflation swaps and other inputs. SF Fed's Cyclical and Acyclical Core PCE Inflation series (Mahedy-Shapiro methodology) decomposes core PCE into cyclical and acyclical components - useful for asking whether remaining inflation is policy-responsive or structural.

Underlying-measure historical interpretation. The Trimmed Mean PCE has the cleanest historical record of identifying the underlying trend with minimal monthly noise; the Cleveland Fed Median CPI has the cleanest record on a monthly basis but tends to lag turning points by 2–4 months; the Atlanta Fed Sticky CPI provides a longer-horizon trend view. Bryan-Meyer (2002, 2010) and Dolmas (2005) are the foundational methodology papers.

Goods-vs-services-vs-shelter decomposition has become the standard analytical frame post-2022. Goods PCE inflation went deeply negative through 2023 (deflation as supply chains normalized and energy prices fell). Services-ex-housing (supercore) was the sticky piece. Housing services peaked late and then decelerated through 2024. Different policy implications: goods deflation reflects supply normalization, not policy success; supercore disinflation reflects labor-market loosening, which is the policy-relevant channel.

Foreign comparison context: the ECB targets 2% headline HICP (Harmonized Index of Consumer Prices) in the euro area - core HICP excluding food and energy is monitored but is not the target; BOE targets CPI in the UK; BOJ targets CPI ex-fresh-food in Japan. Cross-country inflation comparison requires methodology alignment - HICP excludes owner-occupied housing (a major gap the ECB has been working to close); UK CPI methodology differs from CPIH (which includes owner-occupied housing); Japan's CPI methodology differs in food-component definition.

For a one-shot dashboard, the best free sources are: (1) SF Fed Cyclical and Acyclical Core PCE Inflation at https://www.frbsf.org/research-and-insights/data-and-indicators/cyclical-and-acyclical-core-pce-inflation/; (2) Atlanta Fed Underlying Inflation Dashboard at https://www.atlantafed.org/research/inflationproject; (3) Cleveland Fed Inflation Nowcasting and Median/Trimmed CPI page; (4) NY Fed Multivariate Core Trend (MCT) inflation measure - model-based estimate of underlying inflation trend.

Research literature for methodology: Bryan & Meyer (2010, Cleveland Fed Economic Commentary 2010-2) on Sticky-Price CPI; Dolmas (Dallas Fed Staff Paper 2005) on Trimmed Mean PCE; Bryan-Cecchetti (1994) on Median CPI rationale; Stock-Watson on Cyclical vs Acyclical decomposition; Stock-Watson (2007, "Why Has U.S. Inflation Become Harder to Forecast?") and Fuhrer (2010) on inflation persistence. The IMF's Consumer Price Index Manual and BEA's NIPA Handbook (Chapter 5 for PCE) are the comprehensive primary references.

Real-time vintage considerations: monthly PCE and CPI releases are revised. Annual revisions can substantially alter the historical record. For policy-decision-vintage analysis, use FRED's ALFRED archival vintage feature to retrieve the data that was actually available at each FOMC meeting. Modern empirical work increasingly uses real-time vintage data to avoid hindsight bias.

A practical analytical workflow for an inflation update: (1) read the Core PCE M/M and 3-month annualized rates; (2) check the goods vs services vs shelter decomposition; (3) compare Trimmed Mean PCE, Median CPI, Sticky CPI for the underlying-trend confirmation; (4) check 5Y5Y forward breakeven for market expectations; (5) check Michigan and NY Fed SCE for consumer expectations; (6) check Atlanta Fed Wage Growth Tracker or ECI for wage-growth context; (7) read the Fed's most recent Speech-of-Record on inflation framing.