PCE vs CPI: A 65-Year Comparison of the Two Main US Inflation Measures (1960-2026)

Latest observation

The January 2026 data point places the current headline gap in the negative tail of the historical distribution.

Key research findings

1. The 46 bp wedge is structural, not transitory. Across 792 monthly observations from January 1960 through January 2026, the mean CPI−PCE gap is +45.8 bp and the median is +39.7 bp. The standard deviation is 62.2 bp.
2. CPI runs above PCE 4 times out of 5. In 643 of 792 months (81.2%), CPI YoY exceeds PCE YoY. The reverse occurs in 149 months (18.8%), almost all of them clustered in three short windows.
3. Wide divergences are rare but recurring. The gap exceeds 100 bp in absolute terms in 13.3% of months (105 of 792). The 90th percentile of |gap| is 117 bp, the 95th is 153 bp, and the 99th is 263 bp.
4. The Great Inflation produced the widest gap. The 29-month stretch from September 1978 to January 1981 had an average gap of +235 bp and peaked at +374 bp in June 1980, when CPI YoY hit 14.27% while PCE YoY measured 10.52%.
5. Only one sustained reversal exists. Between May and October 1983 — the disinflation following Volcker’s tightening — PCE inflation ran above CPI for six consecutive months, troughing at −178 bp in August 1983.
6. The two series are 98% correlated. CPI YoY and PCE YoY have a correlation of 0.9827 across the full sample. They move together in direction and magnitude; the gap is a level effect, not a divergence in cycle.

Hero chart: 65 years of CPI YoY vs PCE YoY

The two measures track each other closely in cycle. The vertical distance between the lines — shaded red where CPI runs higher, blue where PCE runs higher — is the gap.

Why the gap exists: three mechanical sources

The CPI−PCE gap is not the result of measurement error in either direction. It is the deterministic consequence of three documented differences in how the two indices are constructed: the index-number formula, the category weights, and the scope of expenditure covered.

1. Formula effect

The CPI uses a modified Laspeyres formula at the upper level of aggregation: it holds the expenditure basket fixed for a defined period and updates the basket on a fixed schedule. The PCE uses a Fisher chain-weighted formula that updates expenditure weights every quarter. When consumers substitute away from goods whose prices rise faster, the Fisher formula captures that substitution; the Laspeyres formula does not. The result is that CPI tends to overstate the rise in living costs relative to a chain-weighted alternative — what the BLS and BEA both refer to as “substitution bias” in fixed-weight indices. Until January 2023 the BLS updated CPI weights biennially using two years of expenditure data; starting with January 2023 indexes, weights are updated annually using a single calendar year of expenditure data lagged two years. McCully, Moyer, and Stewart (2007) estimate that the formula effect accounted for roughly half of the 40 basis points per year by which CPI exceeded PCE inflation over the 2002:Q1-2007:Q2 period.

2. Weight effect

The two indices weight expenditure categories differently because they measure different things. The CPI uses out-of-pocket consumer expenditures only. The PCE uses total consumption, including expenditures made on consumers’ behalf by employers (employer-paid health insurance) and government (Medicare, Medicaid). The two resulting weight schemes differ most visibly in healthcare and shelter:

• Healthcare: medical care commodities and services combined accounted for approximately 8.4% of the CPI as of January 2026 (BLS Table 1). PCE assigns a substantially larger share to healthcare — roughly twice as much, since the PCE basket reflects total US healthcare spending including the portion paid by employers and government, while the CPI reflects only direct household out-of-pocket payments.
• Shelter: shelter (rent, owners’ equivalent rent, lodging, and tenants’ insurance) accounted for approximately 35.6% of the CPI as of January 2026 (BLS Table 1), with OER alone weighted at 26.2%. The corresponding shelter weight in PCE is closer to 16% (rent and OER combined). Shelter inflation has run faster than overall inflation in most years since 2010, so the higher CPI weight pulls the CPI level upward.

3. Scope effect

The PCE basket includes some categories of expenditure the CPI excludes entirely, and vice versa. Rural consumers, financial services measured by FISIM (financial services indirectly measured), and a wider set of nonprofit-institution expenditures all appear in PCE but not in CPI. The two scopes diverge most when financial-services prices or rural-area prices move differently from urban consumer prices.

For the full BEA-BLS technical decomposition, see McCully, C. P., Moyer, B. C., and Stewart, K. J. (2007), “A Reconciliation between the Consumer Price Index and the Personal Consumption Expenditures Price Index”, working paper, U.S. Bureau of Economic Analysis, September 2007.

Episode decomposition: 10 sustained wide-divergence stretches

Defining a “sustained wide-divergence episode” as any period of three or more consecutive months in which the absolute gap exceeds 100 bp, the 65-year sample contains ten such episodes — nine in which CPI ran above PCE, one in which PCE ran above CPI.

Nine episodes share a feature: a positive supply shock (oil, food, post-COVID goods) hitting categories where CPI’s higher weight on out-of-pocket prices amplifies the response relative to PCE’s broader basket. The 1983 episode is the mechanical mirror: a sharp disinflation in those same categories pulled CPI down faster than PCE.

Decade-by-decade analysis

The mean gap has varied across decades, but its sign has not. Every decade in the sample shows a positive mean — CPI above PCE — though the magnitude and dispersion differ markedly.

Three observations stand out. The 1990s show the tightest distribution: every single month of the decade had CPI above PCE, and the standard deviation was 23 bp — less than a third of the 1970s value. The 1980s show the widest spread: a 108 bp standard deviation containing both the +374 bp peak and the −178 bp trough. The 2010s show the lowest mean (+24 bp), reflecting the post-GFC era of low inflation and tighter co-movement between the two series.

Regime classification

Classifying each month into one of four regimes based on the absolute size of the headline gap produces a roughly even distribution across “aligned,” “typical range,” and “notable gap” categories, with “wide divergence” reserved for tail months.

A month-by-month classifier (gap_regime) is included in the master dataset so users can map historical episodes or filter for specific regime conditions.

Where each measure is better

A common framing treats the gap as evidence that one index is “right” and the other “wrong.” That framing does not survive contact with the documentation of either measure. Each index is constructed to answer a different question, and each does some things better than the other.

What PCE captures better

• Healthcare expenditure. Roughly half of US healthcare is paid by employers and government, not by households out of pocket. The PCE basket includes both. For analysis of total healthcare-price dynamics — relevant to wage setting and fiscal projections — PCE is the more complete measure.
• Substitution behavior. When relative prices shift, consumers adjust their basket. The Fisher chain-weighted PCE formula reflects this; the modified-Laspeyres CPI updates its basket on a fixed schedule (annually since January 2023, biennially before that). For periods of large relative-price shocks (energy, post-COVID goods rotation), PCE captures realized purchasing behavior more closely.
• Total consumption coverage. The PCE expenditure scope is wider — financial services measured by FISIM, nonprofit-institution expenditures, rural consumers.

What CPI captures better

• Out-of-pocket cost of living. The CPI is constructed to measure the change in prices that urban consumers actually pay from their own wallets. For household budgeting, wage negotiation, and Social Security cost-of-living adjustments (which are explicitly indexed to CPI-W), the CPI is the relevant measure.
• Shelter price changes. CPI weights shelter at roughly twice the PCE level (≈34% vs ≈16%) and measures rent and owners’ equivalent rent through a direct rental-survey methodology. When shelter is the dominant inflation driver, CPI is the more sensitive monitor.
• Timeliness and frequency. CPI is released roughly two weeks earlier each month than the PCE price index and is the underlying input for most financial-market inflation derivatives (TIPS breakevens, CPI swaps).

The two measures coexist in the US statistical system because they address different uses. The persistent 46 bp gap is the price of having both.

Current state: the January 2026 reversal

January 2026 marks one of the rare months in which PCE inflation exceeds CPI inflation. Headline PCE YoY came in at 2.83%, headline CPI YoY at 2.39% — a −44 bp gap. The reading places the current month in the lower tail of the historical distribution: only 18.8% of months since 1960 show PCE above CPI, and most cluster in three short windows (1983, brief stretches in 2002-2003, and a handful of post-2010 readings).

The driver of the current reversal is visible in the category-level data: shelter CPI has decelerated faster than overall PCE inflation through late 2025, while the PCE basket has been more exposed to healthcare-price pressure. The pattern is consistent with the mechanical logic — when shelter disinflation outpaces healthcare disinflation, the CPI−PCE gap turns negative. An illustration of it is found in how input costs reach the CPI through copper. The structural reading is developed in the myths surrounding inflation.

Whether the reversal persists into a sustained episode (defined as 3+ months of |gap| > 100 bp) cannot be inferred from a single observation. Of the 105 wide-divergence months in the sample, only six reached gaps below −100 bp, all in mid-1983. The current −44 bp reading remains well inside the typical range and substantially above the historical minimum.

For comparison, the core gap stands at −60 bp in March 2026 (Core PCE 3.20% vs Core CPI 2.60%), broadly consistent with the headline reversal and similarly inside historical norms.

Five historical turning points

1. June 1980 — the peak (+374 bp). CPI YoY reached 14.27% while PCE YoY measured 10.52%. The widest reading in the 65-year sample, driven by the second oil shock interacting with CPI’s heavier energy weight at the time and its Laspeyres formula amplifying substitution costs.
2. August 1983 — the trough (−178 bp). The disinflation that followed Volcker’s tightening produced a sharp drop in CPI (2.46% YoY) faster than PCE (4.24% YoY). The only sustained reversal in the entire sample; lasted six months before the gap returned to positive territory.
3. October 1990 — Gulf War oil shock pushes gap above 100 bp again. The 1990-91 episode (11 months, peak +138 bp) was milder than 1970s episodes but marked the first wide divergence since 1983.
4. June 2008 — the last pre-COVID wide divergence. Five months of |gap| > 100 bp tied to the oil price spike to $147/bbl. Between 2009 and 2021 — twelve years — no sustained wide-divergence episode occurred, the longest “calm” stretch in the sample.
5. May 2022 — the post-COVID peak (+174 bp). The 2021-22 episode (12 months, average +147 bp) was the first sustained wide divergence in over a decade. The pattern matched prior supply-shock episodes: CPI’s heavier weight on volatile food and energy categories combined with rapidly rising shelter components.

Methodology and reproduction

Data sources

• CPIAUCSL — Consumer Price Index for All Urban Consumers: All Items in U.S. City Average, seasonally adjusted, monthly, Index 1982-1984=100. Source: U.S. Bureau of Labor Statistics, retrieved via FRED.
• PCEPI — Personal Consumption Expenditures: Chain-type Price Index, seasonally adjusted, monthly, Index 2017=100. Source: U.S. Bureau of Economic Analysis, retrieved via FRED.
• CPILFESL — Consumer Price Index for All Urban Consumers: All Items Less Food and Energy in U.S. City Average, seasonally adjusted. Source: BLS via FRED.
• PCEPILFE — Personal Consumption Expenditures Excluding Food and Energy (Core PCE), Chain-type Price Index, seasonally adjusted. Source: BEA via FRED.

Construction

Year-over-year inflation rates are computed as (index_t / index_{t-12} − 1) × 100. The headline gap is defined as cpi_yoy_pct − pcepi_yoy_pct, expressed in basis points (multiplied by 100). The same construction is applied to core (CPILFESL vs PCEPILFE).

Sustained wide-divergence episodes are identified algorithmically as runs of three or more consecutive months in which the absolute headline gap exceeds 100 bp.

Reproduction code (Python)

import pandas as pd
from pandas_datareader import data as pdr

# Pull series from FRED
cpi  = pdr.DataReader('CPIAUCSL', 'fred', '1959-01-01', '2026-05-01')
pce  = pdr.DataReader('PCEPI',    'fred', '1959-01-01', '2026-05-01')

df = cpi.join(pce, how='outer')
df['cpi_yoy'] = df['CPIAUCSL'].pct_change(12) * 100
df['pce_yoy'] = df['PCEPI'].pct_change(12) * 100
df['gap_bp']  = (df['cpi_yoy'] - df['pce_yoy']) * 100

# Long-run mean: ~46 bp
print(df['gap_bp'].dropna().mean())

Limitations

• Basket revisions and methodology changes. Both BLS and BEA periodically revise their methodologies and weight schedules. The BLS introduced geometric-mean aggregation at the lower CPI level in 1999, reducing the formula-effect contribution. In January 2023, the BLS switched from biennial to annual CPI weight updates. Pre-2023 readings are not strictly comparable to post-2023 readings in a methodologically uniform sense.
• Owners’ equivalent rent lag. The OER series in CPI has a documented lag relative to spot rental markets. The Richmond Fed (2023) finds that year-over-year OER growth correlates most strongly with private rents lagged approximately 12 months, and the Cleveland Fed New Tenant Rent Index leads official CPI rent by roughly a year. During periods of rapid shelter-price change, the CPI−PCE gap can widen mechanically before either index reflects realized rents.
• Hedonic adjustment differences. The BLS and BEA apply hedonic quality adjustments to different categories and using different methodologies. The net effect on the gap is small in most periods but can be meaningful when technology-product prices are moving rapidly.
• October 2025 missing observation. The October 2025 CPI release was delayed by the federal government shutdown of late 2025. The dataset shows a missing value for that month; year-over-year calculations bridging that month warrant additional caution.
• Sample-period dependence. The 46 bp average is sensitive to the 29-month Great Inflation episode (1978-81), which contributes an outsized share of the variance. Excluding 1978-82 reduces the long-run mean by roughly 8 bp; readers studying post-Volcker dynamics may prefer the 1983-2026 sub-sample.

Frequently asked questions

Download the dataset

Suggested citation: Eco3min Research (2026). “PCE vs CPI: A 65-Year Comparison of the Two Main US Inflation Measures (1960-2026).” Study #16. Retrieved from https://eco3min.fr/en/pce-vs-cpi-inflation-gap/. CC BY 4.0.

Sources and references

• U.S. Bureau of Labor Statistics. Consumer Price Index Detailed Report. Series CPIAUCSL, CPILFESL. Retrieved via FRED (St. Louis Fed), May 2026.
• U.S. Bureau of Economic Analysis. Personal Consumption Expenditures Price Index. Series PCEPI, PCEPILFE. Retrieved via FRED (St. Louis Fed), May 2026.
• McCully, C. P., Moyer, B. C., and Stewart, K. J. (2007). A Reconciliation between the Consumer Price Index and the Personal Consumption Expenditures Price Index. Working paper, U.S. Bureau of Economic Analysis, September 2007. Available at https://www.bea.gov/sites/default/files/papers/P2007-4.pdf.
• Federal Reserve Board. Statement on Longer-Run Goals and Monetary Policy Strategy. First adopted January 2012, revised in August 2020 and August 2025, most recently reaffirmed January 27, 2026.
• Reis, R. (2021). Losing the Inflation Anchor. Brookings Papers on Economic Activity, Fall 2021.

Related Eco3min research

• Study #14 — US CPI Inflation History: Six Inflation Regimes Since 1913
• Study #9 — Fed Funds Rate and the Taylor Rule (1955-2026)
• Study #7 — Real Returns on US Treasury Bills (1928-2026)
• Study #27 — HY Credit Spreads Lead Equity Peaks