FOUR NUMBERS FOR ONE MONTH OF PRICES — 3.7%, 3.3%, 2.3%, AND 54% — ON SEPTEMBER 30 THE GOVERNMENT REBUILDS THE INDEX AND 0.3 GOES AWAY
No price falls. No family pays less. The yardstick just gets shorter.
One month of prices. Four answers. The government’s headline number for July says 3.7%. Take out food and gas and it says 3.3%. The Dallas Fed throws out the biggest movers at both ends, weights what is left by what people actually spend, and gets 2.3%. Bloomberg counts categories instead of dollars and finds about 54% of them running hotter than 3%, up 16 points since early 2025. Nobody here is lying. They are measuring different things and calling all of it inflation. Then there is September 30. That day the BEA rebuilds three pieces of the index — money management fees, legal work, software — on better price data, and applies the new method back to 2021. Roughly 0.3 points of measured core inflation disappears. Goldman figures May’s 3.4% becomes about 3.2%. No price falls. No family pays less. The yardstick just gets shorter, and the Fed lands closer to its target than it was the day before. That is the argument for hard money written into a release calendar. The people being graded write the test. Bitcoin’s supply schedule was set in 2009 by someone who then walked away from it. It has never been revised, and there is no September 30 on which it could be.
KEY RECEIPTS
July 2026 PCE, released by the Bureau of Economic Analysis on Aug 26 2026: headline +3.7% over 12 months against a 3.6% forecast, core +3.3% matching forecast. Both +0.2% on the month.
Dallas Fed Trimmed Mean PCE for the same month: 2.3% over 12 months, 2.3% annualized over six months, 2.2% annualized on the month. The trimmed mean drops the largest movers at both ends and weights the rest by spending.
Bloomberg Economics, charted from BEA data and posted by The Kobeissi Letter: about 54% of PCE components up more than 3% over 12 months, near the highest share in three years, up about 16 percentage points since Q1 2025.
The 54% and the 2.3% do not contradict each other. One counts how many categories are hot; the other weights by how many dollars are spent in them. A lot of small categories can run hot while the biggest-weight ones do not.
BEA methodology change lands in the Sep 30 2026 release, applied back to 2021. Three categories: portfolio management and investment advice, legal services, computer software and accessories. More of the input shifts to Producer Price Index data.
Expected effect on measured core PCE: about 0.3 percentage points lower in total. Goldman Sachs estimates May 2026 core revises from 3.4% to roughly 3.2%; JPMorgan estimates roughly 3.3%. The three categories are about 4% of core PCE but carry outsized weight in the calculation.
Who profits: The Kobeissi Letter is a paid subscription newsletter and sells urgency. It is relaying a Bloomberg chart, not producing the data. Separately, a lower measured number flatters whoever is being measured against a 2% target. Neither point makes the underlying figures wrong — the BEA change is a real data improvement, not a trick.
Note on timing: the July data is eight days old. What is new is the breadth read and the September 30 rewrite sitting in front of it. Next PCE release: Sep 30 2026.
Sources: bea.gov PCE price index release (Aug 26 2026) + dallasfed.org Trimmed Mean PCE (Jul 2026).
Dallas Fed Trimmed Mean PCE for the same month: 2.3% over 12 months, 2.3% annualized over six months, 2.2% annualized on the month. The trimmed mean drops the largest movers at both ends and weights the rest by spending.
Bloomberg Economics, charted from BEA data and posted by The Kobeissi Letter: about 54% of PCE components up more than 3% over 12 months, near the highest share in three years, up about 16 percentage points since Q1 2025.
The 54% and the 2.3% do not contradict each other. One counts how many categories are hot; the other weights by how many dollars are spent in them. A lot of small categories can run hot while the biggest-weight ones do not.
BEA methodology change lands in the Sep 30 2026 release, applied back to 2021. Three categories: portfolio management and investment advice, legal services, computer software and accessories. More of the input shifts to Producer Price Index data.
Expected effect on measured core PCE: about 0.3 percentage points lower in total. Goldman Sachs estimates May 2026 core revises from 3.4% to roughly 3.2%; JPMorgan estimates roughly 3.3%. The three categories are about 4% of core PCE but carry outsized weight in the calculation.
Who profits: The Kobeissi Letter is a paid subscription newsletter and sells urgency. It is relaying a Bloomberg chart, not producing the data. Separately, a lower measured number flatters whoever is being measured against a 2% target. Neither point makes the underlying figures wrong — the BEA change is a real data improvement, not a trick.
Note on timing: the July data is eight days old. What is new is the breadth read and the September 30 rewrite sitting in front of it. Next PCE release: Sep 30 2026.
Sources: bea.gov PCE price index release (Aug 26 2026) + dallasfed.org Trimmed Mean PCE (Jul 2026).
A promise you keep rewriting isn’t a promise.
The cap is still twenty-one million.
READ THE COVERAGE →
The cap is still twenty-one million.
Bureau of Economic Analysis · July 2026 PCE, released Aug 26 2026 · Also: Dallas Fed Trimmed Mean PCE