October CPI, priced before BLS prints it.
Eighteen slices of the consumer basket, each forecast from prices anyone can watch: pump prices, wholesale used-car auctions, natural gas and jet fuel, and the slow drift of rents. Every slice is weighted the way BLS weights it, then summed into the number on the tag.
What September 2026 told us
The last print against the consensus and the model's pre-release call, and what drove it.
| First print | Now | Consensus | Model | Key Data call | Surprise | |
|---|---|---|---|---|---|---|
| Headline CPI | +0.65% | +0.65% | +0.6% | +0.56% | +0.59% | +0.05 pt |
| Core CPI | +0.33% | +0.33% | +0.2% | +0.20% | +0.21% | +0.13 pt |
- September 2026 CPI rose +0.65% (core +0.33%) against a consensus of +0.6% (core +0.2%).
- The biggest movers in basis points of the headline: gasoline & fuel oil (+34.8 bp), owners' equivalent rent (+9.1 bp), other core services (+4.6 bp).
- Core is running 3.4% annualized over three months and 2.6% over twelve; headline 3.7% year over year.
- The model had called core at +0.20% and headline at +0.56% before the release.
What drove it
Contribution to the headline, basis points.
Analyst notes
Energy did the damage, but core is the story. Headline CPI rose 0.65%, almost exactly what pump prices implied: gasoline and fuel oil alone added about 35 basis points. Strip that out and core still rose 0.33%, a full tenth above the 0.2% consensus.
The three-month pace has turned up. Core is running 3.4% annualized over three months against 2.6% over twelve. Part of the year-over-year figure is still flattered by the October 2025 shutdown, when BLS carried shelter prices forward, so the shorter window is the cleaner read.
Breadth, not one item. Owners' equivalent rent (+0.35%), other core services and hotels all contributed, which makes this harder to dismiss as noise than a single-category spike.
What to watch in October. Pump prices have flattened since mid-September, so the energy boost should fade to a small seasonally adjusted gain; the model puts headline at about 0.23% and core at 0.22%. The risk is on the core side if services momentum persists.
Where the print should land
BLS rounds to one decimal. These are the odds of each rounded outcome, built from the model's own recent miss size, with fat tails for surprise months, plus the uncertainty from the weeks of pump prices still to come.
Headline CPI
Core CPI
The basket, at true scale
Each tile's area is its share of consumer spending. Color shows the change the model expects this month. Shelter alone is over a third of the index, which is why rents set the floor and gasoline sets the swing.
What moves the headline
Contribution in basis points (weight × expected change). Ten basis points is 0.1 on the headline.
How the tag adds up
Expected change and contribution by the four BLS special aggregates.
Open price signals
These public prices lead their CPI counterparts. Weeks already observed are locked. Weeks still to come start at the latest reading; change them to test a scenario and every number on the page updates.
Track record
A real-time backtest: for every month since 2012, the model is refit using only data that existed before that month, then asked to call the print. COVID lockdown months and the 2025 shutdown gap are excluded from scoring.
View every month as a table
Against Wall Street and the Fed
The consensus is the median economist forecast published before each release; the Cleveland Fed publishes a model nowcast. Both are judged against the first number BLS prints, so the fair test scores both forecasts on that first print. Since BLS later revises seasonal factors, the model, which learns from today's revised data, also gets a second score on the revised numbers.
The Key Data call
Once Wall Street's consensus is published, the model's view is combined with it: the call starts from the consensus and moves toward the model by an amount learned from past releases, also weighing last month's surprise and revision. It is used only while it has beaten the consensus over the previous three years, scored out of sample; otherwise the call is the consensus.
The call for October 2026 appears here once the consensus is published, usually a few days before the release.
Headline
| Period | Releases | Model miss | Key Data call miss | Consensus miss | Cleveland Fed miss | Exact rounded print: model / call / consensus | Model closer / further | |
|---|---|---|---|---|---|---|---|---|
| Since 2023 | 43 | 0.075 | 0.070 | 0.074 | 0.092 | 44% / 47% / 47% | 20 / 23 | Call ahead |
| Since 2019 | 87 | 0.113 | 0.100 | 0.100 | 0.123 | 34% / 34% / 36% | 38 / 49 | Even |
| Since 2014 | 140 | 0.101 | 0.092 | 0.093 | 0.102 | 36% / 38% / 39% | 63 / 77 | Even |
Core
| Period | Releases | Model miss | Key Data call miss | Consensus miss | Cleveland Fed miss | Exact rounded print: model / call / consensus | Model closer / further | |
|---|---|---|---|---|---|---|---|---|
| Since 2023 | 43 | 0.061 | 0.067 | 0.071 | 0.077 | 60% / 40% / 40% | 22 / 21 | Model ahead |
| Since 2019 | 87 | 0.100 | 0.100 | 0.101 | 0.117 | 43% / 34% / 32% | 42 / 45 | Call ahead |
| Since 2014 | 140 | 0.086 | 0.088 | 0.088 | 0.098 | 44% / 37% / 36% | 71 / 69 | Model ahead |
Every component
The model's call for each slice next to its last three prints. Type your own number in the last column to override a slice; clear it to return to the model.