Core CPI (ex Food/Energy) Forecast 2026

Core CPI (excluding food and energy) has a better track record than headline because the volatile components are stripped out. Forecasters have still missed turning points: the 2021-2022 acceleration from 1.6% to 6.6% (September 2022 peak) was under-predicted by every major bank; the 2023-2024 disinflation from peak to 3.2% was correctly captured on direction.

Regime-conditional models on core CPI achieve approximately 65% directional accuracy. The print is highly persistent month-over-month, which makes mean-reversion forecasts work better than turning-point forecasts.

Core CPI is a slow-moving regime variable. Trends persist for multiple quarters before shifting. Goldilocks regimes anchor core CPI in 2-2.5% range; stagflation maps to 4%+; deflation maps to sub-1%.

The April 2026 setup has core CPI at 2.6% (down from peak but still above 2% target). The supercore measure (services ex-shelter, the Fed's most-watched metric) is at approximately 4%, materially above target. The gap between core CPI and supercore captures the disinflation challenge: shelter is mean-reverting while supercore services prices are sticky.

Three structural issues drive core CPI forecast errors. First, supercore services (services ex-shelter) is dominated by wages plus services-margin dynamics. Wage growth above 4% (currently 4.0% on Atlanta Fed Wage Tracker) is inconsistent with 2% inflation; the regime model captures the level but not the persistence. Persistent supercore prints near 4% are the dominant reason core CPI has been stuck above 2.5% since mid-2024 despite shelter normalization.

Second, shelter mean-reversion timing is uncertain. The OER methodology lag means shelter inflation continues falling even after market rents stabilize, which creates a mechanical disinflation that ends abruptly when the lag normalizes. The 2024-2025 shelter glide path was relatively smooth; the next leg (2026 onward) depends on whether market rents have re-accelerated, which the BLS print won't capture for another 12-18 months.

Third, idiosyncratic categories (used cars, airfares, medical services) produce noise that swamps signal in any single month. Core CPI ex-volatile-categories (the trimmed-mean PCE analogue from the Cleveland Fed) is a cleaner trend indicator but isn't the print the Fed nominally targets.

For core CPI, watch the supercore services component (the Fed's primary focus) and the shelter component (the largest weight). Supercore stickiness above 3% argues for delayed cuts; shelter normalization below 4% argues for faster disinflation. The two together form the highest-resolution real-time read on whether the disinflation regime is intact or stalling.

The cleanest cross-check is the gap between core CPI and core PCE. Core PCE typically runs 30-50bp below core CPI because of methodology differences (chain-weighted, hospital weights). When the gap widens above 50bp, core CPI is overstating inflation pressure relative to the Fed's preferred gauge; when it narrows below 30bp, core CPI is more representative.

The 68% band on core CPI is the tightest of any inflation series because of the high persistence and low monthly variance. Treat single-month surprises as noise unless they confirm a multi-month trend.