The simple version

A report on household spending comes out and two outlets quote two different growth rates for the same month. Neither of them is wrong. They are reading different lines of the same table.

The release carries a headline total and three narrower figures. Each narrower one removes a different set of categories, so the month can read two different ways depending on which line you pick.

The figure people call core is narrower again, and it is the one the release does not contain. Whoever quoted it built it from the category table underneath.

The numbers

• Headline line, named in the release as "Retail & food services, total": $773,947 million for August 2026 against $764,462 million for July, seasonally adjusted (Census Bureau, Advance Monthly Retail Trade Report for August 2026, released September 16, 2026, Table 1)
• "Total (excl. motor vehicle & parts)": $631,568 million against $622,897 million (same table)
• "Total (excl. gasoline stations)": $711,644 million against $704,007 million (same table)
• "Total (excl. motor vehicle & parts & gasoline stations)": $569,265 million against $562,442 million (same table)
• The four categories the commonly quoted core version removes, August against July: motor vehicle & parts dealers $142,379m and $141,565m; gasoline stations $62,303m and $60,455m; building material & garden eq. & supplies dealers $42,227m and $42,318m; food services & drinking places $105,070m and $103,824m (same table)
• The exclusion set for the core version, as described in a Federal Reserve Bank of Richmond Macro Minute post dated May 30, 2023: automobiles, gasoline, building materials and food services. That post states that its views are the author's and not necessarily those of the Richmond Fed or the Federal Reserve System
• Neither the phrase "core retail sales" nor the phrase "control group" appears in the August 2026 release (checked in the release text, October 10, 2026)

Why one report holds four different growth rates

The narrower lines exist because two categories move on their own logic. Vehicle purchases and fuel sales turn on price and timing rather than on how much households are buying overall, so the release offers versions with each of them taken out.

The core version goes further and removes two more. Many building supply purchases are tied to homebuilding, which the national accounts treat as housing investment rather than household consumption, and the release itself reports food services separately from retail.

What is left is meant to track the goods part of household consumption. That is the reason analysts watch it, and also the reason it has to be assembled by hand.

The actual math on 1.08 and 1.39

Divide each line by its July figure. The headline rose 1.24%. The two lines that each take out one volatile category rose 1.39% and 1.08%, and the line that takes out both rose 1.21%.

The core version has to be built. Take the four categories off the headline and August gives $421,968 million against July's $416,300 million, a rise of 1.36%.

That subtraction can be checked against the document. Removing vehicles alone reproduces its published $631,568 million exactly, and taking out the second one as well reproduces its published $569,265 million, so the method matches the release before it goes past what the release prints.

What this means

If a spending figure surprises you, find out which line it came from before you conclude anything from it. For August 2026 the published lines spanned 1.08% to 1.39% for the same month.

The core version is neither a trick nor an official statistic. It is a convention, computed by whoever is quoting it, from a table anyone can download.

It also means two people can report different core figures for one month and both be arithmetically right, if they remove slightly different categories.

What this is NOT

This article explains what the label core retail sales refers to, and which figures a monthly government release does and does not publish, using the Advance Monthly Retail Trade Report for August 2026, released September 16, 2026, as the example. It is not a forecast of household spending, of any coming release, or of what a future month will show. It does not say which version of the figure is the right one to watch, and it is not advice about your own spending or saving. The exclusion set described here comes from a Federal Reserve Bank of Richmond Macro Minute post that states its views are the author's and not necessarily those of the Richmond Fed or the Federal Reserve System, so it is reported here as a description of a market convention and not as an official definition. No statistical agency publishes this figure. The subtraction producing $421,968 million and $416,300 million, the growth rates computed from the release's lines, and the observation that the method reproduces two of the release's own subtotals are all our arithmetic on its published figures. The advance estimates are themselves drawn from a subsample and are revised in later reports. Census released revised estimates for this period on September 28, 2026, and the figures here are the advance vintage as first published on September 16, not the revised ones. This is not a political endorsement or criticism of anyone.

Sources

• U.S. Census Bureau, "Advance Monthly Sales for Retail and Food Services, August 2026," released September 16, 2026, Table 1 and Explanatory Notes
• Federal Reserve Bank of Richmond, Macro Minute, May 30, 2023 (describes the core retail sales exclusion set; carries an author-views disclaimer)