Truflation and official US CPI
Truflation was 3.08% on 17 October, and official CPI was 3.7% in the year to September. The model estimate for October CPI, which BLS releases on Tuesday 10 November, is 4.0%, with an 80% interval of 3.8% to 4.2%.
Snapshot of 18 October 2026. Data through 17 Oct 2026 for Truflation and September 2026 for CPI.
Loading charts...
Truflation and official CPI, year-on-year %
Both series on one scale. CPI is plotted at the middle of its reference month and Truflation daily; grey bands are NBER recessions. October 2025 CPI was not published. Drag across the chart to zoom and double-click to reset.
Gap between Truflation and CPI, percentage points
Bars above zero mean Truflation is higher. Shaded periods are divergence episodes, a gap of at least 1 pp for three months or more. The two indexes differ in weights, coverage and the treatment of housing.
Rolling correlation over 24 months
Correlation of monthly changes and of levels over the 24 most recent monthly pairs, for the selected CPI measure.
October 2026 CPI: estimates by model with 80% intervals
Forecast date 2026-10-18; Truflation data through 2026-10-16. Intervals are the 10th to 90th percentiles of each model's last 36 out-of-sample errors. The highlighted model has the lowest RMSE since January 2022.
What moves the estimate away from the last print
Starts from the last official print and adds the estimated effect of the change in Truflation and in gasoline prices since that month.
Out-of-sample accuracy, 2015 to 2026
Every model is refitted each month on CPI prints published before the forecast date, and all forecasts are made on the same day of the month.
Forecast error by day of the month
Early in the month the previous CPI print is often not yet published, and Truflation adds the most there.
Estimates against outcomes since 2019
Each point is a forecast made with the data available on day 18 of the reference month.
Accuracy when official CPI is delayed
Simulated interruption of official data: the last available CPI print is 1 to 6 months old. Updating that print with the change in Truflation roughly halves the error of carrying it forward.
Error metrics, pp of the 12-month rate
DM p: Diebold-Mariano test against the random walk with the Harvey-Leybourne-Newbold correction. Direction hit: share of months with the correct sign of change, when the actual move is at least 0.05 pp.
Why the levels differ: gap to headline CPI and to CPI less shelter
Where the two lines diverge, the shelter component explains that part of the gap. Weights and energy exposure also differ.
CPI components and Truflation, year-on-year %
US regular gasoline, USD per gallon
EIA weekly retail price.
Brent crude, USD per barrel
EIA daily spot price.
Monthly panel
Every series used in the charts, one row per month. Sort, filter and export to CSV.
Method notes
- Truflation publishes a daily year-on-year rate. Its monthly value is the average of the daily rates in that month.
- The comparison defaults to CPI-U all items, SA (CPIAUCSL). The nowcast targets the NSA 12-month rate that BLS publishes; it is not revised when seasonal factors are updated.
- Backtests are point in time: a CPI print is used only after its BLS release date, Truflation two days after its observation date and EIA gasoline three days after. Truflation's history before its December 2021 launch is back-filled by the provider.
- 80% intervals are the 10th to 90th percentiles of each model's last 36 out-of-sample errors. The model table reports how often outcomes fell inside them.
- BLS will not publish October 2026 12-month changes because October 2025 was not collected. The month-on-month SA estimate uses a geometric interpolation of the missing October 2025 index between September and November 2025.