Skip to main content

Inflation (CPI, PCE)

Updated Jun 24, 2026 at 2:35pm

Research Draft High 1,265 words

Inflation is the rate at which the general price level of goods and services rises over time, eroding the purchasing power of money. The United States publishes two headline gauges of consumer inflation: the Consumer Price Index (CPI), released monthly by the Bureau of Labor Statistics (BLS), and the Personal Consumption Expenditures (PCE) price index, released monthly by the Bureau of Economic Analysis (BEA). The core tension is that these two measures of "the same thing" are built differently — different baskets, weights, formulas, and scope — and they routinely disagree by a few tenths of a percentage point. That gap matters enormously, because CPI is the number markets trade on the morning of release, while PCE is the number the Federal Reserve targets when setting interest rates.

How it's calculated / formed

CPI tracks the out-of-pocket spending of urban consumers (CPI-U covers roughly the urban population). The BLS prices a fixed basket of goods and services and combines them with a Laspeyres-type formula — weights based on a base period, updated annually (the BLS moved to annual weight updating, from biennial, starting with 2023 data). Because the basket is comparatively fixed, CPI does not quickly capture consumers substituting away from items whose prices rise.

PCE measures consumption from the business/seller side (national-accounts data), so its scope is broader: it includes spending made on behalf of households — notably employer- and government-paid healthcare (e.g., Medicare, employer insurance contributions) — which CPI excludes. PCE uses a chained Fisher-Ideal formula with weights that update essentially every period, so it explicitly captures the substitution effect.

Two structural differences drive most of the divergence, with shelter the largest single line item behind both:

  • Weight (formula) effect — PCE's frequently-updated chained weights tend to run lower than CPI's fixed-basket weights.
  • Scope effect — the broader PCE basket (especially heavy healthcare weighting via third-party payments) versus CPI.
  • Shelter is the largest single category divergence: per Cleveland Fed and other analyses, shelter (rent and owners' equivalent rent) carries roughly a third of CPI (~33–36%) but only ~15–18% of PCE — close to double the weight in CPI. This is why CPI is far more sensitive to housing.

Core versions of both strip out food and energy (volatile, supply-driven) to reveal the underlying trend. Core PCE is the Fed's headline focus.

How it's used in practice

For markets, the CPI report is one of the highest-impact scheduled releases of the month. Traders care less about the absolute level than about the surprise — actual versus consensus expectations. A hot surprise raises expected Fed tightening, lifting yields and pressuring rate-sensitive equities (long-duration growth/tech especially); a cool surprise does the reverse. Research from the Federal Reserve and academics documented that S&P 500 and cross-asset sensitivity to CPI news rose sharply during the 2021–2023 inflation surge (Fed FEDS working papers).

For policy, the FOMC formally prefers PCE (adopted as its reference in 2000; the explicit 2% longer-run target — defined on PCE — was announced in January 2012). The Fed's stated reasons: PCE reduces substitution bias, is revised comprehensively for methodology, and has broader scope. So the practical workflow for a macro-aware participant is: trade the CPI print on the day, but read it as a leading read on PCE, then wait for the PCE release (typically ~2 weeks later) for the number that actually drives the rate path.

Analysts also lean on alternative core measures to filter noise: the Cleveland Fed median CPI, the Dallas Fed Trimmed Mean PCE, and the Atlanta Fed sticky-price CPI. These drop or down-weight outlier categories each month, often giving a cleaner read on the persistent trend than headline or even standard core.

Adoption, debate & evidence

CPI and PCE are the two most-watched inflation statistics in the U.S., and their methodologies are well-established and broadly accepted; this is not a contested indicator in the way a technical pattern is. The genuine debates are about measurement nuance:

  • CPI runs hotter than PCE. Per the Cleveland Fed, since 2000 annual CPI inflation has averaged roughly 0.4 percentage points higher (~0.39 pp) than PCE, with the same gap holding for core CPI vs core PCE. Over decades that compounds materially — relevant because Social Security COLAs and TIPS use CPI, while the Fed targets PCE.
  • Shelter lag. The CPI shelter component lags market rents by roughly a year because it surveys the stock of existing leases, not just new ones. NBER and Fed research documents this; it explains why housing inflation kept rising into early 2023 after market rents had already cooled, and why headline CPI can look "sticky" purely as a measurement artifact.
  • Substitution bias in CPI (the Boskin Commission, 1996, estimated CPI overstated cost-of-living change by ~1.1 pp/year at the time) drove ongoing BLS methodology improvements and partly motivated the Fed's preference for PCE.

The honest summary: both are good, neither is "true inflation." They answer slightly different questions (out-of-pocket cost-of-living vs. total consumption deflator).

Strengths & limitations

Strengths: timely (monthly), long history, transparent methodology, and — for CPI — direct relevance to indexed contracts and to short-term market positioning. PCE's chained, broad design makes it the better measure of the economy-wide consumption price trend.

Limitations / #1 misuse: treating CPI and PCE as interchangeable, or reacting to a single noisy headline print. Energy and food swings, the shelter lag, and one-off category spikes routinely distort a single month. The disciplined read uses core, watches the 3-month and 6-month annualized trend (not just year-over-year), and cross-checks against trimmed-mean/median measures before concluding the trend has changed. A second common error is forgetting that the Fed acts on PCE, so an alarming CPI shelter print may not move policy as much as a headline-reading trader expects.

Sources

Flags: the "CPI ~0.4 pp > PCE since 2000" (~0.39 pp, Cleveland Fed) and "shelter ≈ 1/3 of CPI vs ~15–18% of PCE" figures are well-established across Cleveland Fed/BLS/BEA sources, but exact values drift with annual reweighting and the chosen sample window. The Boskin best-estimate ~1.1 pp/year figure (range 0.8–1.6) is the December 1996 estimate; subsequent BLS reforms have narrowed the bias.