A major consumer price category falls 50% from its peak
Will BLS report any of the eight CPI-U major group indexes (food and beverages, housing, apparel, transportation, medical care, recreation, education and communication, other goods and services) at 50% or more below its previous peak?
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
This is very unlikely. All eight consumer price groups are at or within 1% of record highs, inflation is 3.7% and the Federal Reserve is raising rates. No group has ever lost half its value within five years. Since 1950 the largest drop was 23%, in transportation costs in 2008. What would most change this forecast is the start of a sustained, deep general fall in prices, for example from an extreme economic shock driven by AI.
Where things stand
I downloaded BLS's official CPI-U flat file cu.data.2.Summaries (https://download.bls.gov/pub/time.series/cu/cu.data.2.Summaries, updated 10/14/2026 with September 2026 data) and computed, for each of the eight NSA U.S. city average series, the all-time peak, the latest value, and the YES threshold (half the peak):
| Group | Peak (month) | Sep 2026 | YES needs ≤ | Last month that low |
|---|---|---|---|---|
| Food & beverages (SAF) | 347.662 (Sep 2026) | 347.662 | 173.8 | Jun 2001 |
| Housing (SAH) | 361.392 (Sep 2026) | 361.392 | 180.7 | May 2002 |
| Apparel (SAA) | 140.043 (Sep 2026) | 140.043 | 70.0 | Jul 1974 |
| Transportation (SAT) | 298.409 (May 2026) | 295.862 | 149.2 | Feb 2002 |
| Medical care (SAM) | 593.781 (Jul 2026) | 593.651 | 296.9 | Jun 2003 |
| Recreation (SAR) | 145.854 (Sep 2026) | 145.854 | 72.9 | never (series starts 1993 at about 89.5) |
| Educ. & comm. (SAE) | 150.903 (Sep 2026) | 150.903 | 75.5 | never (starts 1993 at about 84.3) |
| Other goods & svcs (SAG) | 610.165 (Sep 2026) | 610.165 | 305.1 | Jun 2004 |
Six groups set new records in September 2026, and the other two are within 1% of their peaks. Prices are rising, not falling. Headline CPI is up 3.7% from a year earlier, gasoline is up 34.2% because of the U.S.–Iran war, tariffs are pushing apparel up, and the Fed raised rates in September with more hikes priced in (CNBC, 2026-10-14; BLS TED, 2026-10-16).
Base rate
Across the full BLS history of these series (back to 1914/1935/1967/1993), only once did any of them fall to half its previous peak. Apparel did so in June 1931, eleven years after its June 1920 wartime peak, under the gold standard. The largest decline within any five-year window was 41.3% (apparel, 1920–24). Since 1950 the largest was 22.6% (transportation, July–December 2008, driven by the gasoline crash). Food, housing, medical care, recreation, education and communication, and other goods and services never fell more than 3% from a prior peak. So the reference rate for a 50% fall within five years under fiat money and an inflation-targeting Fed is zero in about 75 years.
Paths to YES and why each is remote
- Energy collapse: gasoline is only 2.9% of CPI against 16.3% for transportation (BLS relative importance, December 2025). Even gasoline going to zero would cut transportation by roughly 20–25%, not 50%.
- Policy making services free: education is about 44.5% of education and communication, so free tuition and childcare alone would still fall short of 50%. Health insurance is only about 10.6% of medical care.
- Transformative-AI or depression-scale deflation: this would require a fall in the general price level the Fed has never allowed and would fight hard. Housing, food and vehicles are dominated by land, labor and physical inputs that are unlikely to halve in nominal terms by 2031. This is the main tail, and it grows later in the window.
- Rebasing: this is excluded by the rules. BLS's May 12, 2026 rebasing to December 2024 = 100 covered only selected sub-series with extreme levels (tobacco, hospital services, TVs, IT and others), not major groups, and BLS recalculates history on the new base.
- Data error: a 50% error in a major-group index would also distort the all-items index. Past BLS corrections (for example the electricity corrections of 2024 and 2026) were small.
Hazard model
The near-term hazard is essentially zero: even the fastest modern collapse, 22.6% in five months, is less than half the needed move, and everything is at its peak. I assumed per-quarter hazards of about 0.001% in late 2026, rising to about 0.023% by late 2031 to capture AI-transformation, depression and data-error tails. That gives a cumulative 0.08% by end-2027, 0.15% by end-2030 and 0.23% by end-2031.
Strongest case against
This question belongs to a set built around transformative AI, and a sufficiently radical economic break could defeat historical regularities. Strain on BLS (budget-driven collection cuts in 2025, the missing October 2025 data, expert panels on missing Consumer Expenditure data) slightly raises the risk of errors. Even so, a nominal halving of a broad group needs years of deep general deflation, so I would still take far better than 99:1 odds on NO.
Sources: BLS flat files (download.bls.gov/pub/time.series/cu/, 2026-10-14); BLS rebasing notice and list (bls.gov/cpi/notices/2026/rebasing.htm; bls.gov/cpi/additional-resources/rebased-series.htm, 2026-05-12); BLS 2028 geographic revision notice (2026-07-27); BLS relative importance, December 2025 (bls.gov/cpi/tables/relative-importance/2025.htm); CNBC September CPI report (2026-10-14); BLS TED on gasoline (2026-10-16).
Resolution criteria
Question. By the last day of each quarter, has BLS reported any of the eight CPI-U major group indexes at 50% or more below its previous peak?
Definitions.
- Indexes: the monthly, not seasonally adjusted, U.S. city average Consumer Price Index for All Urban Consumers series listed below; annual averages are not observations for this question.
| Major group | BLS series ID |
|---|---|
| Food and beverages | CUUR0000SAF |
| Housing | CUUR0000SAH |
| Apparel | CUUR0000SAA |
| Transportation | CUUR0000SAT |
| Medical care | CUUR0000SAM |
| Recreation | CUUR0000SAR |
| Education and communication | CUUR0000SAE |
| Other goods and services | CUUR0000SAG |
- Previous peak: the highest published monthly index level in the same series before the observation month, using its entire available BLS history, including months before the question opened. Use the historical values available as of the qualifying publication, incorporating any BLS corrections already published.
- 50% or more below: the observation's index level is less than or equal to 0.5 times its previous peak, calculated from published index levels without rounding the resulting percentage to reach the threshold.
- Compare levels on the same reference base. A rebasing alone does not count. A renamed or recoded series counts only if BLS identifies it as a continuation of the same major-group series and supplies comparable history or an official linking method. Routine BLS basket, weight, and quality-adjustment changes do not disqualify the series.
Resolves YES if BLS publishes at least one qualifying monthly observation meeting the threshold. The Observation period convention applies: the observation month must end after the opening date. A single month suffices; no sustained decline or attribution to AI is required. BLS need not explicitly describe the decline from the peak if its published levels establish it. A correction that first establishes the threshold counts on the correction's publication date; later corrections do not reverse an established YES under the Resolution window convention.
Resolves NO for a quarter if no qualifying publication has appeared by that quarter-end. Regional or metropolitan indexes, seasonally adjusted series, CPI-W, chained CPI, subcomponents such as televisions or computers, and aggregates other than the eight listed groups do not count. A halving of an inflation rate, expenditure weight, or price relative to wages or the all-items CPI does not count. Missing or discontinued observations are not treated as zero.
Resolving source. U.S. Bureau of Labor Statistics only: its CPI database, Public Data API, downloadable CPI-U time-series files, or official CPI releases and corrections reporting the listed series. Series identifiers follow BLS's CPI series ID codes and item-code directory. Apply the Resolution window and Observation period conventions and publication-date rule A3; this named-source requirement is narrower than Public confirmation (A2).