Inflation

Reviewed 11 August 2026 · Sourced from the BLS, the BEA, the Federal Reserve and the Census Bureau

What inflation actually measures

Your rent went up at renewal. Ground beef costs more than it did last spring. That is what most people mean by inflation. The official statistic means something narrower: a sustained rise in the general price level — the same sentence as a sustained fall in what a dollar buys.

The word general carries the weight. One price rising on a bad harvest or a new tariff is a relative price change, and it signals where to send more supply. Inflation is the whole level drifting up together — a statement about money, not about beef.

Two things fall out of that arithmetic. A year-over-year rate can drop purely because the month it is measured against was unusually high — nothing has to happen this month. And a falling rate does not mean falling prices. It means prices are still rising, slower. Prices only fall when the rate goes negative, which is deflation and a different problem.

How the basket is weighted, and why shelter runs it

CPI prices a basket meant to stand in for what an urban household buys — a population that, in the BLS's words, “constitutes over 90 percent of the U.S. population.” Each category is weighted by its share of spending, and the weights decide everything: a category at 30% of spending moves the index thirty times as hard as one at 1%.

Shelter is not a component of the U.S. CPI. It is the component.

The BLS does not use house prices, because buying a house counts as investment rather than consumption. For the 65.0% (±0.5) of households who owned their home in the second quarter of 2026, it uses owners' equivalent rent — an estimate of what the home would rent for. That reflects the average across all existing leases, so it turns months after the housing market does.

Since January 2023 the BLS updates the weights annually, from one calendar year of Consumer Expenditure Survey data on a two-year lag — which is why the December 2025 table is labeled 2024 weights. Before 2023 the update came every two years, from two years of data.

Two official measures of the same month, disagreeing

Two agencies publish inflation and they disagree. CPI comes from the Bureau of Labor Statistics. The PCE price index comes from the Bureau of Economic Analysis, inside Personal Income and Outlays. CPI lands first — for June 2026, CPI on 14 July and PCE on 30 July, a 16-day gap.

BEA names four reasons they diverge. The formula effect — PCE uses the Fisher-Ideal formula, CPI a modified Laspeyres. The weight effect. The scope effect: “PCE measures spending by and on behalf of the personal sector, which includes both households and nonprofit institutions serving households; the CPI measures out-of-pocket spending by households.” And other effects, covering seasonal adjustment and residual differences. Scope is why medical care weighs more in PCE, which counts care an employer or a government program pays for; shelter weighs more in CPI.

What does not follow is a direction. BEA asserts no general rule about which measure runs higher, and June 2026 has PCE above CPI on both headline and core. The gap is a mechanism, not a constant you can subtract.

One difference outranks the rest: the Fed's longer-run objective is 2 percent on the annual change in the PCE price index — headline, not core, and not CPI. Holding a CPI print up against that 2% measures with the wrong ruler, and it is done constantly.

Core, and the measures built because core has limits

“Core” is market shorthand. The agencies say all items less food and energy (BLS) and PCE price index excluding food and energy (BEA). Food and energy swing on weather, harvests and geopolitics, and a central bank cannot make it rain, so core is an attempt to see the persistent trend underneath. It is not a claim that food is free — and it understates a long energy shock, which eventually feeds into everything else.

Because core has that failure mode, regional Federal Reserve banks publish alternatives that cut the data differently. Cleveland's median CPI is the one-month rate of the component whose expenditure weight sits at the 50th percentile of price changes. Its 16% trimmed-mean CPI averages the components whose weights fall between the 8th and 92nd percentiles. Atlanta's sticky-price CPI tracks only components whose prices adjust infrequently — the cutoff is not published, so nobody outside the bank should be quoting a threshold.

There is also core services excluding shelter, informally called supercore, on the reasoning that those prices are mostly wages. The three-way split behind it — core goods, housing services, and core services other than housing — is Chair Jerome H. Powell's, from his November 2022 Brookings speech. Powell never used the word supercore. The market attached it afterwards.

What the market thinks inflation will be

Expectations are not a mood. They are priced daily in the Treasury market. Inflation-protected securities adjust for realized inflation; ordinary Treasuries do not. Subtract one yield from the other at the same maturity and you have the breakeven inflation rate.

Three readings show how far apart the sources sit. The 10-year breakeven was 2.29% on 10 August 2026. The 5-year, 5-year forward rate — expected inflation on average over the five years beginning five years from today, far enough out that a current shock has washed through — was 2.33% the same day. The New York Fed's Survey of Consumer Expectations, run monthly on a rotating panel of roughly 1,300 household heads, put the median one-year-ahead expectation at 3.6% in July 2026. Households sit well above markets. They usually do.

Nominal is what you are quoted. Real is what you keep.

A savings account paying 4% while inflation runs 6% loses you purchasing power every month it sits there, however positive the statement looks. This is the one piece of arithmetic worth carrying off this page.

Applied to policy, the same subtraction tells you whether money is actually tight. As of 30 July 2026 the federal funds target range is 3-1/2 to 3-3/4 percent, with interest on reserve balances at 3.65%. Headline PCE inflation is 3.7%. On the shortcut convention the real policy rate is roughly zero — neither restrictive nor stimulative — behind a nominal rate that sounds high. The 10-year TIPS real yield was 2.40% on 7 August 2026.

One further point, offered as reasoning rather than as a sourced figure, because no agency publishes it as a rule: nominal rates cannot fall far below zero, since people would hold cash instead. So when inflation goes negative the real rate is pushed up even at a zero policy rate, and policy tightens exactly when the economy needs the opposite. The related claim that resetting unanchored expectations has historically required deliberately inducing a recession is an interpretation, not a measurement.

What is legally tied to the number

Inflation is not only a description. Federal payments and contracts are wired to specific index series, and rarely the one in the headline.

Social Security. The cost-of-living adjustment is set on CPI-W, the Consumer Price Index for Urban Wage Earners and Clerical Workers — not the CPI-U everyone quotes — comparing a third-quarter average against the third-quarter average of the last year a COLA took effect. CPI-W covers about 30% of the U.S. population against CPI-U's over 90%. The 2026 COLA is 2.8%.

TIPS. They index to the non-seasonally-adjusted CPI-U on a three-month lag, and Treasury's offering circular states it flatly: “The Ref CPI for the first day of any calendar month is the CPI for the third preceding calendar month.” The protection is real and arrives a quarter late by design.

So before assuming the headline rate applies to a raise, a benefit letter or a bond prospectus, find which index the document names. CPI-U, CPI-W and headline PCE are three different numbers for the same month.

Where the number has holes

An honest page about inflation has to say where the statistic has holes.

Two months are missing. Because of the lapse in appropriations, the BLS collected no CPI data from 1 October through 12 November 2025. In the agency's words, “Missing CPI data affected October and November 2025 indexes,” and it propagated forward: “Missing October 2025 data also affected April 2026 rent and owner's equivalent rent indexes.” The Cleveland Fed had to interpolate October 2025 to keep its median and trimmed-mean series running. Any twelve-month comparison spanning that window rests partly on prices nobody collected.

Every figure here has a shelf life. The July 2026 CPI is due at 8:30 a.m. Eastern on 12 August 2026, and it restates the twelve-month rate. That is why each number above carries its as-of date. For live readings rather than dated ones, the economic pillar pulls CPI, core, PCE and the breakevens straight from the source.

The figures on this page are checked against the source that publishes them, and dated. Published rates move after the release named above — the linked source always carries the current number. This page explains a term; it does not recommend a product.