Browse the library

Economics · Economic indicators

Core inflation: why food and energy are left out

What core inflation is, why it drops food and energy, how much monthly CPI noise it removes, and how to annualize and read core prints.

Intermediate 8 min read By Finabase Research Updated Oct 30, 2026

On this page

What core inflation leaves out, and why
How much noise core removes
Reading a monthly core print
Core over long periods
Other trend measures
Pitfalls and limitations
Core and its neighbours
How Finabase does it
Frequently asked questions
Summary

Core inflation is the inflation rate with food and energy prices taken out. In the US the two standard versions are core CPI, from the Bureau of Labor Statistics, and core PCE, from the Bureau of Economic Analysis. Food and energy are what people buy most often and notice most, so leaving them out can look like a trick. It is not meant as one. The idea is to separate the persistent part of inflation, which policy can influence and which tends to continue, from price swings that often reverse within months.

The difference is not academic. In September 2026, BLS reported headline CPI up 3.7% over the year and core CPI up 2.6%, a gap of more than a full point, almost all of it from energy prices that were 20.4% higher than a year earlier. Which of those numbers describes "the" inflation rate depends on the question you are asking.

This page explains what core leaves out and why, shows on a decade of monthly data how much noise it removes, explains how to annualize monthly core readings, and covers the alternative trend measures some Federal Reserve banks publish.

What core inflation leaves out, and why

Food and energy prices respond to things that monetary policy cannot steer: weather and harvests, crop diseases, refinery outages, wars and the production decisions of oil exporters. These shocks are often large and often temporary. A spike in gasoline prices that reverses within a few months moves headline inflation up and then down without saying much about where inflation is heading a year out.

The excluded categories are not small. In BLS’s December 2025 relative-importance table:

Category	Share of CPI-U, %
Food	13.7
— Food at home	8.3
— Food away from home	5.4
Energy	6.4
— Motor fuel	3.0
All items less food and energy (core)	79.9

What remains in core is mostly services, led by shelter (35.6% of the whole index), medical care and transportation services, plus goods such as vehicles and apparel. These prices tend to move slowly and, once they start rising faster, tend to keep doing so. That persistence is exactly what makes core useful for judging the trend.

How much noise core removes

Month-over-month % change in CPI-U all items (headline) and all items less food and energy (core), seasonally adjusted, 2015 to September 2026. Dashed lines at 0% and 0.165% (≈2% a year). BLS published no October 2025 index, so there is no monthly change for October or November 2025; the lines run flat across that gap. Source: BLS via FRED (CPIAUCSL, CPILFESL).

The chart plots every monthly change since 2015. The headline line swings well above and below the core line, which hugs a narrower band. Measured over the 139 consecutive month pairs from 2015 to September 2026:

Statistic, monthly % change	Headline CPI	Core CPI
Standard deviation	0.28	0.17
Months with a decline	18	5
Lowest month	−0.79	−0.49
Highest month	+1.26	+0.81

Headline’s extremes come from energy: the lowest month was April 2020, when the energy index fell 9.5% in a single month, and the highest was June 2022, at the peak of the energy surge. Core’s extremes came in the same pandemic years but were smaller.

2026 gives a clean example of why the distinction matters. In March 2026 headline CPI rose 0.87% in a single month while core rose 0.20%. In June headline fell 0.42% while core was essentially flat (−0.02%). Read one at a time, those headline months suggest first an inflation spike and then deflation. Core, at 0.2% to 0.4% a month for most of the year, told a steadier story.

Reading a monthly core print

From a monthly core change to an annual rate

annualized = (1 + monthly change)^12 − 1

2% a year   ⇔  (1.02)^(1/12) − 1 = 0.165% a month
0.2% a month →  1.002^12 − 1   = 2.4% a year
0.3% a month →  1.003^12 − 1   = 3.7% a year
0.4% a month →  1.004^12 − 1   = 4.9% a year

So a "0.2" core month is roughly in line with a 2% to 2.5% trend, and a "0.3" or more is clearly above it. Market reactions on CPI day often hinge on that second decimal place.

Single months are still noisy, even for core. A common compromise is to annualize three or six months. For core CPI at September 2026, computed from the seasonally adjusted index:

Window	Core CPI, annualized	Headline CPI, annualized
September alone (1 month)	4.1%	8.1%
June–September (3 months)	3.4%	4.6%
March–September (6 months)	2.9%	3.7%
12 months (BLS, unadjusted)	2.6%	3.7%

The shorter the window, the bigger the swings. The table shows core running a little faster over recent months than over the full year, while the one-month headline figure, 8.1% annualized, is mostly a single month’s energy move.

One data warning: BLS published no October 2025 value for the all-items or core index after a lapse in federal appropriations stopped price collection. When computing changes across that gap, match calendar months: there is no one-month change for October or November 2025, and a November-over-September change covers two months.

Core over long periods

Excluding food and energy does not bias inflation down over time, because those prices rise and fall around the same long-run trend as everything else. From January 2000 to January 2020, headline CPI rose 2.15% a year and core 2.01%. Over 1990–2020 the two were 2.39% and 2.37%. The gap matters over months, not decades.

The months can be long, though. In July 2008 headline CPI inflation was 5.5% and core 2.5%. A year later headline was −2.0% and core 1.5%. Someone steering by headline would have seen a surge and then deflation; core showed a modest slowdown.

Other trend measures

Dropping food and energy is a blunt rule: it removes those categories even when they are calm and keeps every other category even when one of them spikes, as used cars did in 2021. Some Federal Reserve banks publish alternatives:

Median CPI (Federal Reserve Bank of Cleveland): each month, the price change of the component at the 50th percentile of the spending-weighted distribution of price changes.
16% trimmed-mean CPI (Cleveland Fed): a weighted average that drops the components with the top 8% and bottom 8% of price changes.
Trimmed Mean PCE (Federal Reserve Bank of Dallas): a similar trimming applied to the PCE components, calculated from BEA data.

These measures let the data decide each month which prices are outliers, instead of fixing the excluded list in advance.

Pitfalls and limitations

Core is not your cost of living. Food and energy were about a fifth of the CPI basket in December 2025, and a larger share for many lower-income households.
Energy leaks into core. Higher fuel prices raise the cost of airfares, deliveries and many services over time. A long energy shock can push core up after a lag.
Not every shock is food or energy. Core can still be pushed around by one-off moves in other categories, which is why trimmed and median measures exist.
Core CPI is not the Fed’s main gauge. The Fed’s goal is defined on headline PCE inflation, and it watches core PCE closely as a guide to the trend. See PCE inflation: the Fed’s preferred gauge vs the CPI.
Rounding hides a lot. A published 0.3% core month can be 0.25% to 0.349%, which annualizes to between 3.0% and 4.3%.

Core and its neighbours

Measure	Excludes	Published by	Best for
Headline CPI	Nothing	BLS	Cost of living, TIPS, headlines
Core CPI	Food and energy	BLS	Trend in consumer prices
Core PCE	Food and energy	BEA	The trend the Fed watches most
Median CPI	All but the middle price change	Cleveland Fed	Trend, robust to outliers
Trimmed-mean CPI	Top and bottom 8% of price changes	Cleveland Fed	Trend, robust to outliers
Trimmed Mean PCE	Price-change tails	Dallas Fed	Trend on the PCE basket

How Finabase does it

Finabase does not publish CPI or core inflation data, and it has no economic calendar. The chart on this page is ours, built from BLS series downloaded via FRED on October 30, 2026.

The app does show the market prices behind the volatile parts that core removes. The futures page lists energy contracts, including WTI crude oil, Brent, natural gas and RBOB gasoline, and farm contracts such as wheat, corn and soybeans. The Treasury page shows how bond yields move after an inflation release, and the energy sector page shows how oil and gas stocks are trading.

Frequently asked questions

Why do economists exclude food and energy?

Because those prices swing for reasons monetary policy can’t control, such as weather and oil supply, and the swings often reverse. Removing them gives a clearer read on the persistent trend.

Is core inflation lower than headline inflation?

Not systematically. From 2000 to 2019 headline CPI averaged 2.15% a year and core 2.01%, almost the same. Core is lower when food and energy prices are rising fast, as in September 2026, and higher when they are falling, as in 2009.

What monthly core reading is consistent with 2% inflation?

About 0.165% a month, because 1.00165 compounded twelve times is about 1.02. Published to one decimal, that is a "0.2".

Does the Fed target core inflation?

No. The Fed’s 2% goal is written for headline PCE inflation. Fed officials watch core PCE closely because it is a better guide to where headline will settle.

Summary

Core inflation removes food and energy, about a fifth of the CPI basket, to reveal the slower and more persistent trend. Monthly headline CPI is roughly two-thirds noisier than core by standard deviation, and in 2026 it jumped 0.87% in March and fell 0.42% in June while core stayed near 0.2% a month. Over decades the two measures nearly match, so core is a tool for reading the next year, not a way to make inflation look lower. Read monthly prints against the 0.165% a month that equals 2% a year, prefer 3- or 6-month annualized rates over single months, and use trimmed-mean and median measures as cross-checks when one non-energy category is distorting core.