US August 2026 CPI: month-on-month disinflation continues.
The June 2026 print was the cool one, and it was cool on the month rather than on the year. This call tests the same series that carried the signal, month on month, and states up front how short the sourced run behind it is.
Falsifier as filed, verbatim: “US Aug 2026 CPI MoM print >= 0.0% (net cool or flat) fires MET”. Grading note: the line as filed pairs a symbol with a parenthetical that points the other way, since a month-on-month print at or below 0.0 percent is what “net cool or flat” describes while the symbol reads as flat-to-warm. Tempora grades the parenthetical, at or below 0.0 percent, and records the notation conflict here and on the receipt rather than resolving it quietly.
Stated probability: 0.35. Base rate quoted below from a six-print run rather than a frozen cohort. Calibration tier: Lab, signature under test. Reconciliation within 3 days of the BLS release.
The base rate, and how thin it is
The BLS release for June 2026, published on 14 July 2026, put the seasonally adjusted all-items change at minus 0.4 percent on the month, the first monthly decline in five months, with the index less food and energy unchanged on the month and up 2.6 percent over the twelve months. On that run, one of the six prints through June sat at or below zero, which is a recent-run frequency near 17 percent. Core was flat on the month, so a flat test clears on core even where headline does not.
That number is quoted with its weakness attached. Six prints is a small sample with a wide interval around it. Tempora holds no frozen month-on-month cohort for US CPI, so no full-history frequency appears here, and the stated probability of 0.35 sits above the six-month run because the June composition was broad rather than one-line, with core flat alongside the headline decline. Two points of tilt on a thin base is the whole of the claim.
Coverage of the June print also carried a longer-history claim that the monthly decline was the first since 2000. That sits badly against the April 2020 monthly decline, so Tempora does not use the 2000 framing and quotes the five-month statement from the release-cycle record instead.
Method: lesson applied
The lesson applied here is metric-select for month-on-month rather than year-on-year, logged as M3 after USCPIJUN. That call asked for headline CPI year-on-year at or below 2.5 percent for June 2026 and graded FAILED at 3.5 percent, while the month-on-month print came in at minus 0.4 percent and core at 2.6 percent. The reconciliation recorded the cause as metric-selection: a year-on-year headline threshold is loaded with base effects from the prior year, so it can move against the underlying trajectory that the reading was about. This call moves the test onto the series that carried the signal.
The limitation of that move is equally plain. Month-on-month prints are noisier than annual ones, single months swing on gasoline and airfares, and the seasonally adjusted series is revised when BLS updates seasonal factors. Cleaner alignment to the signal comes with a louder monthly signal-to-noise problem, and a Lab entry is the right surface for that trade.
Test condition and falsifier
MET: the August 2026 all-items seasonally adjusted one-month percent change prints at or below 0.0 percent in the BLS release. FAILED: any print above 0.0 percent. The falsifier as filed reads “US Aug 2026 CPI MoM print >= 0.0% (net cool or flat) fires MET” and the grading note above governs the direction.
The live branches out are named now. Gasoline and energy reversing after a soft summer would push the headline positive on its own. Shelter has been the slow component in this cycle and a single firm shelter month can carry the index. Tariff pass-through into goods prices is the third branch, and a BLS revision to seasonal factors can move a marginal print across the line after the fact. On the quoted run, the failure branch is the more common one, which is why the stated probability sits at 0.35 rather than above a half.
Chart-side note
This instance asserts no transit or period mechanism. The dated fact sheet that governs astronomical claims for this batch was unavailable at publication, and Tempora's rule is that chart-side dates come from the dated sheet or they do not appear. The call therefore ships on its measurable side alone.
That is the honest position for a Lab entry in any case. Lab signatures are under test rather than established, the surface exists so that experimental readings grade in public before they reach the flagship Tracker, and a call with no chart-side layer attached still earns or loses a verdict on the same window.
Limitations, stated
- The base rate comes from six prints, so its confidence interval is wide enough to contain the stated probability.
- Seasonally adjusted month-on-month figures are revised; a marginal print can change side after publication.
- The falsifier as filed carries a notation conflict, recorded above and on the receipt.
- No chart-side mechanism is claimed for this instance, so nothing here tests the framework layer.
- The BLS release date for August data is expected in the second week of September 2026 and is not yet confirmed on this page; reconciliation follows the release, whenever it lands.
Sources
- BLS, Consumer Price Index Summary, June 2026 results: bls.gov/news.release/cpi.nr0.htm. All items down 0.4 percent on the month seasonally adjusted, up 3.5 percent over the year unadjusted, index less food and energy unchanged on the month and up 2.6 percent over the year.
- BLS, June 2026 CPI release archive: bls.gov/news.release/archives/cpi_07142026.htm. Release date 14 July 2026.
- CNBC, June 2026 CPI coverage, 14 July 2026: cnbc.com. First monthly decline in five months, print below street forecasts.
- BLS release schedule: bls.gov/schedule. July 2026 CPI published 12 August 2026, with August data expected in the second week of September 2026.
Frequently asked
What is the US August 2026 CPI call?
The all-items seasonally adjusted month-on-month change in the BLS Consumer Price Index for August 2026 prints at or below 0.0 percent. At or below fires MET; above fires FAILED. Stated probability 0.35 against a recent-run frequency near 17 percent from the six prints through June 2026.
Why month-on-month rather than year-on-year?
Lesson M3, applied after USCPIJUN. That call used a headline year-on-year threshold and graded FAILED at 3.5 percent while the month-on-month print was minus 0.4 percent and core was 2.6 percent. Year-on-year headline thresholds are loaded with prior-year base effects. Month-on-month tracks the disinflation timing question more directly, at the cost of a noisier series.
What is the base rate and how solid is it?
One of the six prints through June 2026 sat at or below zero on the seasonally adjusted all-items series, a frequency near 17 percent. It is a six-print run rather than a frozen cohort, so it is quoted with that weakness attached. Tempora holds no full-history month-on-month cohort for US CPI.
What is the chart-side mechanism?
None is claimed for this instance. The dated fact sheet governing astronomical claims for this batch was unavailable at publication and Tempora does not assert transit dates without it. The call grades on its measurable side alone.
When does Tempora reconcile?
Within 3 days of the BLS release of the August 2026 CPI report, expected in the second week of September 2026, on the public Lab surface with the release figure quoted.
Read next
Lab-tier forward call published by Tempora Research. Lab entries are experimental signatures under test on dated windows. Methodology reproducible against the public engine using Swiss Ephemeris with True Pushya Paksha ayanamsa (PVRN Rao). Internal audit log maintained. This article does not constitute investment, financial, legal, medical or professional advice. First published 30 July 2026 by Tempora Research.