Core CPI vs Core PCE

Core CPI (CPILFESL on FRED) is published by the Bureau of Labor Statistics from the Consumer Price Index for Urban Consumers, dropping food and energy and using a Laspeyres-style fixed-basket framework that updates weights every two years. Core PCE (PCEPILFE) is constructed by the BEA from Personal Consumption Expenditures and uses a Fisher-ideal chain index that updates weights monthly. The two indices are built off different source surveys: CPI relies on the Consumer Expenditure Survey of household spending, while PCE relies on retail-sales receipts and business surveys, which is why PCE captures Medicare and employer-paid health insurance that never appears on a household receipt.

The Fed has formally targeted Core PCE since 2000 and explicitly committed to 2% in January 2012 at the FOMC. Core CPI is what households recognize because BLS releases it on the second Tuesday of the month, weeks before BEA releases the corresponding PCE. As of the BLS release covering March 2026, Core CPI ran 2.6% year-over-year while Core PCE printed 3.2% in the same window, the rare configuration where Core PCE exceeded Core CPI.

The single biggest mechanical difference is shelter: housing carries roughly 33% of the Core CPI basket but only about 15% of Core PCE. When BLS Owners' Equivalent Rent runs hot, as it did from mid-2022 through mid-2024, Core CPI prints meaningfully above Core PCE because the same shelter signal hits CPI with double the weight. The November 2022 peak gap reached 142 basis points, with Core CPI at 6.0% and Core PCE at 4.7%.

Healthcare runs the other way. PCE incorporates employer-sponsored health insurance, Medicare, and Medicaid via BEA's national accounts, and healthcare is the largest single category in Core PCE at roughly 17%. CPI healthcare reflects only what consumers pay out-of-pocket and weights at about 7%. PCE healthcare prices are also smoothed by Medicare administered-pricing rules that lag market rates, which is why physician-services inflation can show very different prints in the two series within the same quarter. The August 2024 BLS-vs-BEA divergence on physician services was 90 basis points, almost entirely driven by Medicare reimbursement methodology.

Both the Cleveland Fed Median CPI and the Dallas Fed Trimmed-Mean PCE were designed to strip the noise that drives the Core CPI-Core PCE gap. The Cleveland Fed Median CPI, published monthly alongside the BLS release, takes the weighted-median price change across all CPI components and consistently runs closer to Core PCE than Core CPI does. As of the March 2026 release, Median CPI printed 3.5% versus Core CPI at 2.6% and Core PCE at 3.2%, a configuration that historically signals services inflation is more entrenched than the headline CPI deceleration suggests.

Dallas Fed Trimmed-Mean PCE excludes the most extreme price changes each month and is the cleanest read on persistent inflation. It tends to lead Core PCE by one to two months at turning points. When Trimmed-Mean PCE turned higher in Q1 2021, it preceded the Core PCE surge by approximately six weeks. The Fed staff explicitly references both measures in the Tealbook, treating them as the institutional check on whether a Core CPI-Core PCE gap reflects signal or methodology.

In 1979-1981, Core CPI peaked at 13.6% in June 1980 while Core PCE peaked at 10.2%, a gap of 340 basis points driven mostly by mortgage-interest-cost treatment that BLS only removed from CPI in 1983. The 1983 methodology change closed roughly 150 basis points of the structural gap overnight.

The 2007-2008 commodity surge produced a smaller divergence because Core excludes food and energy. The 2021-2023 episode was dominated by shelter: Core CPI peaked at 6.6% in September 2022 while Core PCE peaked at 5.6% in February 2022, partly because PCE's chain-weighted methodology adjusts faster when consumers substitute away from rapidly inflating categories. The 2024-2026 episode flipped the historical pattern. Shelter disinflation in CPI pulled Core CPI below Core PCE for the first time since 2021, while services ex-housing in PCE remained sticky, producing the November 2025 inversion that Goldman Sachs and the Cleveland Fed both flagged as a structural shift in the relationship rather than a noise-level deviation.

Core CPI prints on the second Tuesday of the month at 8:30am ET; Core PCE typically prints on the last business day of the month at 8:30am ET. The roughly two-week gap between releases is itself tradeable: macro desks build a Core PCE nowcast from the CPI print plus the PPI components that map into PCE healthcare and financial services, and the residual error on that nowcast moves Treasuries on PCE day. The standard model error is around 8 basis points on month-over-month Core PCE, so any surprise larger than 15bp typically moves 2-year Treasuries 3-5bp on the print.

The September 2024 release illustrated the workflow. After Core CPI printed 3.3%, the consensus PCE nowcast pointed to 2.7% Core PCE; the actual September PCE print landed at 2.65%, and 2-year yields moved less than 2bp despite the headline beat. Practitioners use the nowcast residual, not the headline print, as the trade trigger.

For Fed-policy positioning, Core PCE is the only series that maps directly to the SEP and the dot plot. The September 2025 SEP projected Core PCE at 2.6% for end-2026 versus the 2.0% long-run goal, and that projection drives the policy path. Core CPI matters for TIPS pricing because the entire Treasury Inflation-Protected Securities market accrues on headline CPI, not PCE; the breakeven calculation backed out from TIPS prices is a CPI-implied breakeven, never a PCE-implied one.

For wage-indexed contracts and Social Security cost-of-living adjustments, CPI-W (the urban wage-earner version) is the legal reference. The 2026 COLA was set at 2.5% based on the Q3 2025 average of CPI-W. Households whose spending baskets are heavily weighted toward shelter feel CPI; investors whose returns track Treasuries feel CPI; corporate margins and Fed reaction functions feel PCE. Reading both series is how you avoid mistaking a methodology gap for a regime change.