CPI vs. PCE: why the U.S. has two inflation numbers

Not financial advice. This explainer is for understanding the data, not for making investment decisions.

When people say "inflation," they usually mean one number. The U.S. government publishes two main ones each month, from two agencies, and they don't always agree. In August 2026, the two headline numbers matched at 3.4%, but the "core" versions were 0.6 percentage points apart.

The two gauges

The Consumer Price Index (CPI) comes from the Bureau of Labor Statistics. It tracks the prices urban consumers pay out of pocket. The main version, CPI-U, covers more than 90% of the U.S. population. BLS staff collect prices every month in 75 urban areas from about 6,000 housing units and about 22,000 retail and service establishments (BLS CPI release, Technical Note).

The Personal Consumption Expenditures (PCE) price index comes from the Bureau of Economic Analysis. It covers the prices of goods and services bought by people in the U.S. "or those buying on their behalf" (BEA). It comes out a couple of weeks after the CPI, in BEA's Personal Income and Outlays report.

The Federal Reserve's 2% inflation goal is defined with PCE, not CPI. In its official strategy statement, the Fed says inflation at "2 percent, as measured by the annual change in the price index for personal consumption expenditures" fits its mandate best over the longer run (Federal Reserve, Statement on Longer-Run Goals).

Why they differ

BEA sorts the gap into four buckets (BEA FAQ):

1. Scope. CPI counts what households pay themselves. PCE also counts spending made for them. The big example is health care: PCE includes care paid for by employer insurance, Medicare, and Medicaid, which CPI leaves out (BLS; Cleveland Fed).
2. Weights. CPI weights come mainly from a household survey of what people say they bought. PCE weights come from business surveys of what companies sold (BLS).
3. Formula. CPI uses a modified Laspeyres formula, which mostly holds the basket fixed between updates. PCE uses a Fisher-Ideal formula, which adjusts as people switch to cheaper items when prices rise (BEA; BLS).
4. Other. This covers differences in seasonal adjustment and pricing, plus leftovers.

The weights matter a lot. In CPI, shelter is 35.6% of the basket, energy is 6.4%, and medical care is 8.4% (BLS relative importance, Dec. 2025). PCE gives housing less weight and health care more. So a jump in rents or gasoline tends to move CPI more, and a jump in hospital prices tends to move PCE more. BEA's own worked example from 2006 shows CPI running 0.7 points above PCE. Most of that gap came from CPI's bigger weights on owner housing and gasoline, partly offset by PCE's wider health-care coverage (BEA FAQ).

One more difference: CPI-U is final when it's released. PCE gets revised. BEA's August 2026 release included an annual update that revised estimates back to January 2021 (BEA, Aug. 2026).

The latest numbers (August 2026)

Sources: BLS, Sept. 11, 2026; BEA, Sept. 30, 2026.

Energy drove much of the CPI number. The CPI energy index was up 16.3% from a year earlier, and gasoline was up 27.4%. Gasoline alone made up more than a third of August's monthly CPI increase (BLS). That's why the headline number (3.4%) sits so far above core (2.4%): core strips out food and energy because they swing so much.

The unusual part is that core PCE (3.0%) is running above core CPI (2.4%). Historically it's more often the other way around. BEA publishes a line-by-line reconciliation of the two indexes (NIPA Table 9.1U). We haven't reviewed it for August, so we won't guess at the cause here.

Which one should you watch?

It depends on the question:

• "How much more am I paying at the register?" CPI is closer to out-of-pocket costs, and BLS notes that many contracts and pension plans adjust payments using unadjusted CPI.
• "What is the Fed aiming at?" PCE, because that's how the Fed defines its 2% target.
• "What's the underlying trend?" Look at core versions of both, and read more than one month.

Neither number is "the real" inflation. They measure different things in different ways, and comparing them tells you more than either one alone.

Next releases: CPI for September on Oct. 14, 2026. PCE for September on Oct. 29, 2026.

Not financial advice. Nothing here is a forecast or a recommendation.

---

Sources

• U.S. Bureau of Labor Statistics, "Consumer Price Index — August 2026" (Sept. 11, 2026), including Technical Note. https://www.bls.gov/news.release/cpi.nr0.htm
• BLS, "Relative importance of components in the Consumer Price Indexes, December 2025." https://www.bls.gov/cpi/tables/relative-importance/2025.htm
• BLS, "Differences between the Consumer Price Index and the Personal Consumption Expenditures Price Index" (Beyond the Numbers). https://www.bls.gov/opub/btn/archive/differences-between-the-consumer-price-index-and-the-personal-consumption-expenditures-price-index.pdf
• U.S. Bureau of Economic Analysis, "Personal Income and Outlays, August 2026" (Sept. 30, 2026). https://www.bea.gov/news/2026/personal-income-and-outlays-august-2026
• BEA, "Personal Consumption Expenditures Price Index." https://www.bea.gov/data/personal-consumption-expenditures-price-index
• BEA FAQ, "What accounts for the differences in the PCE price index and the Consumer Price Index?" https://www.bea.gov/help/faq/555
• Federal Reserve, "Statement on Longer-Run Goals and Monetary Policy Strategy." https://www.federalreserve.gov/monetarypolicy/files/FOMC_LongerRunGoals.pdf
• Federal Reserve Bank of Cleveland, "PCE and CPI Inflation: What's the Difference?" (2014). https://www.clevelandfed.org/publications/economic-trends/2014/et-20140417-pce-and-cpi-inflation-difference

Verification notes

• Verified (primary, 2026-10-05): all August 2026 CPI and PCE figures, release dates, CPI coverage and sample sizes, CPI relative-importance weights, the Fed's 2% PCE wording, BEA's four-effect breakdown and its 2006 example, PCE revisions back to Jan. 2021, and the next release dates.
• Not verified with a number: "PCE gives housing less weight and health care more" is stated qualitatively, based on BEA's 2006 decomposition and the BLS and Cleveland Fed explainers. We didn't pull current PCE expenditure shares.
• Not verified: "Historically it's more often the other way around" (core CPI above core PCE) reflects long-run patterns described by BEA and BLS, but we didn't compute a current long-run comparison. Soften or cut it if needed.
• Not explained on purpose: why core PCE exceeded core CPI in August 2026. We didn't review BEA Table 9.1U for that month.
• Not verified: the "couple of weeks after the CPI" timing is based on this month's dates (Sept. 11 vs. Sept. 30) and the coming October dates. It isn't an official rule.
• Time-sensitive: if the September CPI (Oct. 14) or PCE (Oct. 29) is out before publishing, update the table.