Services vs Goods Inflation: The Great Post-2020 Divergence

The post-2020 services-versus-goods divergence: how a single inflation regime split into two sub-regimes operating on different timetables — and why monetary policy struggled to read both at once.

Goods inflation peaked first and rolled over fast; services inflation peaked roughly a year later and stayed sticky for many quarters. The split was visible across the U.S., the U.K. and the euro area, and it explains why the disinflation debate remained alive long after the headline number had fallen.

Treating “inflation” as a single number missed the most diagnostic feature of the post-2021 cycle: the two sub-regimes ran asymmetrically. Reading goods and services separately turns the headline volatility into a structured story about supply chains, wages and the long-cycle of housing.

The split, in dates and numbers

Goods led, then collapsed

U.S. CPI durable goods inflation peaked at +14.3% YoY in March 2022 (BLS), driven by used cars, appliances and electronics caught in supply-chain bottlenecks. The same series fell to +0.6% by mid-2023 and into negative territory (-2.5%) by mid-2024 as supply chains normalized and used-car prices unwound. Non-durable goods (food, energy, household supplies) followed a similar but less extreme pattern: peaking near +13% in mid-2022, declining to +1.5% by 2024. The compression of the durable goods cycle was exceptionally fast — a feature of supply-driven shocks that resolves once the supply constraint clears.

Services lagged, then stuck

U.S. CPI services inflation peaked at +6.5% YoY in February 2023 (BLS) — eleven months after goods peaked. Services excluding shelter and energy peaked at +6.0% YoY in early 2023 and remained above 4% through mid-2024. The persistence reflected wage dynamics in labor-intensive sectors (restaurants, healthcare, education, personal services) and the long lag in the shelter component, which is the largest single services line item.

The cross-region pattern

The euro-area HICP showed a similar pattern with regional variation. Goods inflation in the euro area peaked at +14.6% in October 2022 and fell to -0.3% by mid-2024 (Eurostat). Services HICP peaked at +5.6% in mid-2023 and remained sticky around 4% into 2024. The U.K. picture was more extreme: services CPI peaked at +7.4% in mid-2023 (ONS), the highest among major economies, partly reflecting the UK’s tighter labor market and contractual indexation patterns.

Why the two sub-regimes diverged

Goods: supply chains and the inventory cycle

The 2021 goods inflation was a textbook supply-shock pattern: pandemic-era demand for durables surged just as global shipping capacity collapsed. The Drewry container index peaked at $10,377 per 40-foot container in September 2021 (Drewry), against a 2018-2019 norm of $1,500. Shipping costs fell back to normal by mid-2023 — and goods inflation followed almost mechanically. Inventory cycles amplified the swing: retailers and manufacturers ordered double in 2021-2022, faced excess stock by mid-2023, and discounted aggressively. Goods CPI is, in this sense, a real-time read on supply conditions; exchange-rate transmission compounds the effect for open economies.

Services: wages and shelter as anchors

Services inflation responds more slowly because services pricing is anchored in wage costs and contracts. U.S. average hourly earnings growth peaked at 5.9% YoY in March 2022 (BLS) and declined gradually through 2024 to roughly 4%. The pass-through to services prices runs at a multi-quarter lag. Wage-price spirals describe the channel; the empirical record shows that wage gains transmit to services CPI with a 6-12 month delay in advanced economies. Shelter compounds the lag: shelter inflation, computed via owners’ equivalent rent in the U.S., follows market rents with a 12-18 month lag, which means even a sharp turn in real rental markets takes time to show up in CPI.

Different elasticity to monetary policy

Monetary tightening transmits faster to goods than to services. Higher rates cool durables demand (cars, appliances financed by credit) within quarters; the cooling effect on labor markets and services pricing takes 12 to 24 months. Federal Reserve and ECB tightening cycles in 2022-2023 therefore produced visible goods disinflation before any dent in services inflation appeared — a sequence that fed the public debate about whether central banks were doing enough.

The wage-to-services pass-through in detail

The transmission from wages to services pricing operates through several margins. Direct labor-cost pass-through dominates in labor-intensive services (restaurants, healthcare, personal care) where wages can represent 40-60% of operating costs. Markup-preservation pricing — firms adjusting prices to keep margin ratios stable as costs rise — is the cleanest mechanism. A second margin operates through unit-labor-cost dynamics: even if wage growth moderates, services pricing continues if productivity growth slows, because what matters for margins is the ratio. The 2023-2024 services stickiness reflected both: nominal wage growth around 5% and productivity growth around 1.5% generated unit-labor-cost growth of 3.5% — exactly consistent with the observed services CPI persistence around 4-5%. Reading the wage component without the productivity component produces incomplete forecasts of services pricing dynamics.

The shelter sub-component

Why shelter dominates the services aggregate

U.S. CPI assigns shelter roughly 33% weight in the headline; within services, shelter accounts for over half. Owners’ equivalent rent (OER) — the imputed rental value of owner-occupied housing — alone is about 25% of total CPI. Movement in OER therefore moves headline services CPI mechanically. The OER methodology surveys actual rents in matched neighborhoods and applies the rent change to owned housing — a procedure that introduces the well-documented lag relative to market rents.

The 2022-2024 shelter cycle

Real-time rental-market indices (Apartment List, Zillow Observed Rent Index) showed annual rent growth peaking at +18% YoY in late 2021. CPI shelter peaked at +8.2% YoY in March 2023 — fifteen months later (BLS). By mid-2024, real-time rent indices were near 0% YoY while CPI shelter was still printing +5%, illustrating the lag in reverse. The eventual cooling of CPI shelter in 2024-2026 was therefore predictable from the 2022-2023 rental-market data, but the lag meant that headline CPI looked stickier than the underlying housing cycle.

Cross-region speed differences

The pace at which goods and services disinflated also varied across regions. The U.S. delivered the fastest goods rollover (peak to trough in roughly 18 months) while the U.K. experienced the slowest services disinflation. The euro area sat between, with goods unwinding broadly in line with the U.S. but services tracking somewhere between the U.S. and U.K. paths. The pattern reflected differences in monetary-policy timing, energy-mix shocks and labor-market structures rather than measurement differences alone.

Reading services and goods separately

Goods CPI as a supply-chain read

Goods inflation is the right indicator for monitoring supply-chain pressure, manufacturing capacity utilization and the trade cycle. The New York Fed’s Global Supply Chain Pressure Index, the Baltic Dry Index and the Drewry container index all correlate strongly with goods CPI at quarterly frequency — usually leading goods CPI by one to two quarters.

Services excluding shelter as a wage read

Federal Reserve communications in 2022-2024 increasingly emphasized “supercore” inflation — services excluding shelter and energy — as the cleanest read on wage-driven price pressure. Supercore peaked at +6.5% YoY in early 2023 and remained above 4% well into 2024. The metric is the closest empirical proxy to the Phillips-curve framework: it isolates the inflation component most directly tied to labor-market tightness.

Implications for forecasting

The 2022-2024 forecasting failures (central-bank, market-economist and IMF projections all underestimating inflation persistence) flowed substantially from underestimating the services-shelter persistence and overestimating the speed of pass-through from rate hikes. Reading the components separately would have flagged the asymmetry in real time. Why central banks systematically miss inflation forecasts includes this sub-regime confusion as one of the structural drivers.

The 2024-2026 unwind

Services normalization

By late 2024 and into 2025, U.S. services excluding shelter had cooled to roughly 3.5% YoY, while shelter was decelerating from its 5%+ levels toward the 3% range as the rental-market cooling propagated through OER. The euro area showed a similar but slower normalization: services HICP at 3.7% in mid-2024 versus 5.6% peak. The path of services disinflation is more diagnostic of the durability of the disinflation than the headline number, because services inflation is the component most resistant to reverse acceleration.

The risk of re-acceleration

Services inflation re-accelerates when wage growth re-accelerates, which is most likely in tight labor markets with renewed demand pressure. The 2024-2026 environment combines moderating but still-elevated wage growth (around 4% in the U.S., 3.5% in the euro area) with cooling unit-labor-cost growth as productivity rebounds. The combination is consistent with services inflation continuing to drift toward 3% rather than re-accelerating, but the regime is not yet anchored. A close reading of the 2021-2024 episode documents the unwind in detail. The complete inflation guide develops the regime framework; the four-measure comparison shows how services-versus-goods plays differently in CPI versus PCE; the inflation sub-pillar aggregates monthly readings across this split.

Last updated — 18 May 2026