Why Census Raised August Retail Sales Growth to 1.3% While September Showed 0.8%
The Census Bureau's monthly retail sales figure is a sample-based estimate that is later revised. Here is how the survey is built and what the September 2026 release showed.

Each month the U.S. Census Bureau publishes an advance estimate of retail and food services sales for the month just ended. The estimate covers businesses in two groups that the Census Bureau reports together in its headline total. The Census Bureau says these estimates are released approximately ten business days after the end of the reference month, and they are built from a sample of companies rather than a count of every seller.
The first number is not the final one. The September 2026 estimate, released Oct. 15, 2026, showed a 0.8% monthly gain (±0.4%). In the same release, Census raised the prior month's change to up 1.3% (±0.1%) from up 1.1% (±0.4%). The sections below cover how the total is defined, how the sample is drawn, how the figures are calculated and how earlier figures are revised.
What the Headline Total Includes
The headline total combines two groups of businesses. Retail trade, classified under NAICS 44-45, covers establishments that sell merchandise in small quantities to the general public. Food services, under NAICS subsector 722, covers establishments that prepare meals, snacks and beverages to customer order for eating on or off the premises.
The September 2026 total of $745.0 billion, seasonally adjusted, covers both groups together, according to the Census advance report dated Oct. 15, 2026.
Nonemployer businesses were treated differently in the past. Census says they were previously included through a benchmarking adjustment factor. As of the April 2025 annual revision, they are no longer included in the estimates and have been removed from the time series, to align with the Annual Integrated Economic Survey.
Because the headline combines the two groups, a change in either group is part of the total reported each month.
Who Answers the Survey
The advance estimate is drawn from a sample rather than a count of every seller. Census's advance methodology page describes a sample of about 4,800 companies and employer identification numbers (EINs), selected from a Monthly Retail Trade Survey sample of about 13,000 units.
The Monthly Retail Trade Survey methodology page describes a stratified, one-stage design with 85 industry strata. Each stratum is split into 4, 7, 10 or 13 annual sales size strata, and the largest sales stratum in each industry is selected with certainty.
The advance sample is redesigned and reselected about every two and a half to three years. The advance sample adds no births between redesigns. The Monthly Retail Trade Survey adds new EINs in February, May, August and November and removes deaths at the same time; in that survey, new units are weighted in over three months, and removed units phase out in reverse. Census says births are added to that survey roughly nine months after they begin operation.
The Monthly Retail Trade Survey methodology page says estimates are suppressed when the coefficient of variation is above 30% or the total quantity response rate is below 50%. Sampling weights in that survey range from 1 to 500.
Census publishes its monthly figures as estimates with margins of error. The September 2026 monthly change of 0.8% carries a margin of ±0.4%, according to the Oct. 15, 2026 release.
How the Advance Estimate Is Calculated
Census's advance methodology page says the advance estimates for the most detailed industries are computed with a type of ratio estimator known as the link-relative estimator. The page also says Census does not estimate monthly totals before estimating the month-to-month change.
For nonrespondents, the page says imputation is not performed for most of them. For a limited number of nonresponding companies with influential effects, sales may be estimated from the company's historical performance.
Collection runs on a short schedule. Telephone follow-up begins on the third business day, and collection is completed by the sixth business day. Publication follows about ten business days after the reference month ends.
Seasonal Adjustment and Annual Benchmarking
The published monthly figures are seasonally adjusted. The advance methodology page says the estimates are adjusted for seasonal variation and for holiday and trading-day differences, using the X-13ARIMA-SEATS program with the X-11 filter-based adjustment procedure.
Monthly totals are also benchmarked to the latest annual survey totals, according to the Monthly Retail Trade Survey methodology page. Annual estimates for 2023 and 2024 come from the Annual Integrated Economic Survey, adjusted using the 2022 Economic Census and 2022 published inventories from the Annual Retail Trade Survey. Beginning with the April 25, 2025, benchmark report, estimates were restated to the 2017 NAICS.
Annual estimates for 2022 and earlier come from the Annual Retail Trade Survey for retail sales and from the Service Annual Survey for food service sales, adjusted using the 2022 Economic Census.
Why the July-to-August Change Was Revised
The first published figure for the July-to-August change was up 1.1% (±0.4%). In the September 2026 release, Census shows that change revised to up 1.3% (±0.1%). The revised figure carries a narrower margin than the original.
On Census's sales page, the revised prior-month change sits alongside the new month's figures. For September 2026, the same page reports the monthly change, the year-over-year change and the revision together.
The Monthly Retail Trade Survey page describes two procedures that affect monthly estimates. One is benchmarking monthly totals to the latest available annual survey totals. The other is the addition and removal of EINs in the sample described above, phased in to minimize the effect of births and deaths on month-to-month change estimates.
Reading September 2026 and the Next Release
Census reported that September 2026 sales rose 0.8% (±0.4%) from August to $745.0 billion, seasonally adjusted. Compared with September 2025, sales were up 6.3% (±0.5%), according to the release page dated Oct. 15, 2026.
The October 2026 report is listed for Tuesday, Nov. 17, 2026, at 8:30 a.m. ET. FedRateCalc lists that date and says its dates come from the official Census Bureau release schedule.
The date is a third-party listing.
The Census economic indicators calendar lists 2027 release dates, including Jan. 15, 2027, for December 2026 sales and Dec. 15, 2027, for November 2027 sales.
For readers, the sources show two points. A monthly figure is published with a margin of error, and it can be revised in a later release along with its margin. The September 2026 release reports both the original and revised July-to-August figures.






