US consumer prices fall more than 3% in a year
Will BLS report twelve-month CPI-U inflation below −3% for any month?
Chance by Q4 2031
1%
Cumulative probability
- Latest forecast
| Year | Q1 | Q2 | Q3 | Q4 |
|---|---|---|---|---|
| 2026 | <1% | |||
| 2027 | <1% | <1% | <1% | <1% |
| 2028 | <1% | <1% | <1% | <1% |
| 2029 | <1% | <1% | <1% | <1% |
| 2030 | <1% | <1% | <1% | <1% |
| 2031 | <1% | 1% | 1% | 1% |
Model reasoning
U.S. consumer prices have not fallen more than 3% in a year since 1938. Even the 2008–09 financial crisis, with a record collapse in energy prices, produced only a 2.1% decline. Prices are now rising about 3.5% a year during an oil shock, and rents are sticky, so this would take a deep depression or an unprecedented upheaval. The development that would most change this forecast is a sudden end to the Iran war followed by a severe recession with falling rents.
Where things stand
- Prices are rising, not falling. The August 2026 release (Sept 11, 2026) put the CPI-U 12‑month change at +3.4%. Energy was up 16.3%, core up 2.4% and shelter up 3.0%. In Table 1's July 2026 relative importances, energy is 7.35%, food 13.54%, core 79.11% and shelter 35.34%.
- Next release. September CPI comes out Oct 14. Kalshi and the FactSet consensus expect about +3.6%.
- Background. The Iran war began Feb 28, 2026 and the Strait of Hormuz is still closed. Brent is about $106–108 and gasoline averaged $4.36 in September, up 37.6% from a year earlier. The Fed raised rates in September. Payrolls rose only 29k and unemployment is 4.2%.
- Calendar quirk. There is no October 2025 index, so no 12‑month change can be computed for October 2026.
Base rate (my calculations on BLS flat files)
I used download.bls.gov/pub/time.series/cu/cu.data.1.AllItems (CUUR0000SA0) and cu.data.20.USCommoditiesServicesSpecial (SA0E, SA0L1E), computing year-over-year changes rounded to one decimal.
- No month below −3.0% since 1938. The last was Nov 1938. Since 1947 the low is −2.9% (1949). In the modern era it is −2.1% (July 2009).
- Record component lows. Energy's record 12‑month fall is −28.1% (July 2009). Core's record low is +0.6% (Oct 2010); core has never been negative since 1957.
- Other floating-currency rich economies. Japan's deepest reading was about −2.5% in 2009. My rough count is about 750 country-years since 1950 with no reading below −3%, which implies about 0.1–0.2% a year.
What it would take
- The arithmetic. Use current weights and flat food prices. Even a record energy collapse (−28%) would need core at about −1.25%. A −40% energy collapse, far beyond the record, would still need core at about −0.1%.
- Near-term index thresholds.
- November 2026 (the last reading published in Q4) needs a 6.2% drop from August in three months. The record three-month drop is −3.9% (Dec 2008).
- February 2027 needs a 5.4% drop in six months. The record is −4.0%.
- Mid‑2027 needs about −3.0% over about 12 months, matching or beating the post‑1947 record of −2.87%.
- Ending the oil shock is not enough. If gasoline returned to its pre‑war level, headline inflation would land around 0%, not −3%, because shelter (35% of the basket) and services are sticky.
Paths to YES
- An oil price collapse combined with a depression, including outright falls in rents and core prices. This is the worst case for 2027–28.
- Deflation from a 1930s‑style financial crisis that monetary and fiscal policy fail to stop.
- A transformative‑AI shock, such as mass job losses and service-price falls with the Fed stuck at zero rates, mostly in 2029–31.
- A negligible chance of a BLS error or a methodology change.
Strongest case against my estimate
2026 is unusual: a large oil shock that could reverse sharply, the Fed tightening into a weakening labor market, yields at multi‑decade highs, talk of an AI bubble, and the chance that AI changes price dynamics in ways history doesn't capture. Thailand fell below −3% in 2009 and 2020 through energy price falls plus subsidies. Still, the US basket is dominated by shelter, and the Fed and Congress strongly resist deflation. This keeps me above a pure base rate, but only a little.
Hazard model
Quarterly hazards:
| Period | Per quarter | Per year |
|---|---|---|
| Q4 2026 – Q1 2027 | 0.02% | |
| Q2 – Q4 2027 | 0.03–0.05% | |
| 2028 | 0.05% | 0.2% |
| 2029 | 0.06% | |
| 2030 | 0.07% | |
| 2031 | 0.08% |
Accumulated, this gives about 0.16% by end‑2027 and about 1.2% by end‑2031.
Sources
- BLS CPI August 2026 release, Table 1: https://www.bls.gov/news.release/archives/cpi_09112026.htm and https://www.bls.gov/news.release/cpi.t01.htm
- BLS flat files: https://download.bls.gov/pub/time.series/cu/
- Morningstar, Oct 12, 2026 (consensus 3.6%, Fed hike): https://www.morningstar.com/economy/cpi-forecasts-show-moderated-inflation-september-keeping-fed-hold
- CNBC, Oct 12, 2026 (Kalshi): https://www.cnbc.com/2026/10/12/inflation-will-remain-sticky-in-2026-according-to-prediction-markets.html
- StockTitan, Oct 12, 2026 (gasoline $4.36): https://www.stocktitan.net/articles/september-2026-cpi-inflation-report
- Oil and Iran war: CNBC, Oct 12, 2026; Reuters, Oct 4 and 11, 2026; Britannica "2026 Iran war"
- Jobs: AP and NYT, Oct 2, 2026
Resolution criteria
Question. By the last day of each quarter, has BLS reported twelve-month CPI-U inflation below −3% for any month?
Definitions.
- CPI-U: the Consumer Price Index for All Urban Consumers, all items, U.S. city average, not seasonally adjusted, BLS series CUUR0000SA0.
- Twelve-month inflation: BLS's published unadjusted percentage change from a calendar month one year earlier to the reference month, not an annualized monthly change or an annual-average change.
- Below −3%: strictly less than −3.0%, using the percentage change as BLS reports it. At the current one-decimal reporting precision, −3.1% or lower qualifies; −3.0% does not. Do not substitute a more precise calculation from index levels for the published percentage change.
Resolves YES if a qualifying BLS publication reports this rate below −3% for at least one reference month of September 2026 or later. The Observation period and Resolution window conventions apply. An initial release or a subsequent official correction or revision can qualify; a later revision does not undo a YES. No attribution to AI, productivity growth, abundance, recession, or any other cause is required.
Resolves NO if no qualifying report has been published by the quarter-end. Regional or metropolitan CPI, individual expenditure categories, core CPI, CPI-W, chained CPI, seasonally adjusted changes, forecasts, and observations for August 2026 or earlier are excluded.
Resolving source. U.S. Bureau of Labor Statistics only: the monthly Consumer Price Index news release, specifically Table 1, “All items” row and unadjusted twelve-month percent-change column; its dated archived releases; or an official BLS correction or database publication of the corresponding twelve-month percentage change for CUUR0000SA0. Equivalent BLS tables count if the table numbering or series identifier changes but the measure remains the same. This is a narrower source requirement than Public confirmation (A2). Use the qualifying publication date under Resolution date (A3), not the reference month; database-only changes require a verifiable BLS publication date.