Food Dollar Series Wholesale Share Data: Dataset Structure and Field Coverage

It is the slice of every dollar American consumers spend on food that pays for wholesale trade - distributors' sales forces, warehouses and delivery networks - measured by the USDA ERS Food Dollar Series. In the series' industry group bill, food wholesale is category 8, and for 2024 it totals 6.3 cents per domestic food dollar, or 137,393.8 million dollars across all food spending.

The same bill prices twelve neighboring stages in identical units: farm production, food processing, transportation and storage, food retail and food services among them, so the wholesale number always arrives with its supply-chain context attached. Values are published both as cents per domestic food dollar and as millions of dollars, and the wholesale stage is further decomposed by primary factor - salary and benefits, property income, output taxes and imported inputs.

That decomposition is what separates this series from a headline margin statistic. The 2024 salary-and-benefits line alone shows how much of the wholesale stage is payroll rather than margin, which matters when you are benchmarking a distribution business against the value it adds instead of against its revenue. Because every stage is expressed per dollar of the same denominator, shares can be compared across years without deflating anything first.

1. Food Dollar Nominal Data (XLSX, 9.34 MB) - the wide pivot view, one column block per series.
2. Food Dollar Nominal Data in Machine Readable Format (ZIP, 982 KB) - contains food-dollar-nominal.csv with 126,900 rows plus a readme data dictionary. This is the file pipelines should use.
3. Archived 2011-model files - XLSX/CSV pairs covering nominal and real 1993-2023 series (5,952 rows each).

The machine-readable CSV is linearized - one value per row - so filtering to the wholesale share is a single predicate: keep rows where Series equals Industry group bill, CategoryName equals Food wholesale and ValueTypeDesc equals Total. Documentation sits on the product's Documentation subpage, alongside the underlying Ag-FEDS description, a users' guide and a recorded training webinar.

Account codes follow a fixed pattern worth knowing before you write any parser. Every spending account starts with Xf followed by four digits: the first two identify the channel and the last two are a unique account ID, so Xf1000 is the all-food total and Xf1101 is cereals in the food-at-home channel. The 47 accounts divide into 6 aggregated accounts (food, alcohol and their at-home/away-from-home splits), 28 food-at-home product accounts, 7 food-away-from-home outlet accounts such as full-service restaurants and schools, and 6 alcohol product or outlet accounts. Each account is expressed both in millions of dollars and in cents per domestic food dollar.

Thirteen columns describe every row, and four of them carry the whole selection logic. Year runs 2007-2024 in the current nominal file; Series names one of five bills (Total spending, Marketing bill, Industry group bill, Industry group by primary factor, Primary factor bill); CategoryName carries the stage label, where Food wholesale is category 8; and ValueTypeDesc splits values into Total plus the four primary factors. Unit then tells you whether a value reads as cents per domestic food dollar or millions of dollars.

Field definitions in Datadory's record are flagged verified rather than inferred, which matters because the linearized layout is unusual: analysts accustomed to wide pivot tables sometimes mistake Aggregation values (Total, Type total, Detail) for a hierarchy they are not, and double-count detail rows on top of type totals. Filter on one aggregation level at a time.

The series is annual and national, so treat it as a calibration layer rather than an operational feed. A defensible first pass:

Two structural caveats shape what the model can claim. First, coverage is national totals only - the series publishes no state or sub-state detail, so regional margin gaps need a different source. Second, the primary-factor decomposition lets you test whether a change in the wholesale share is a labor story or a margin story: salary and benefits, property income, output taxes and imported inputs are each reported separately for category 8, in both units.

One caution ERS states explicitly: the current release and the archived 2011-model release are not directly comparable, because the methodology changed when the model was rebuilt on top of BEA input-output tables.

Datadory catalogs 17 pooled records for food distribution work: only 2 are native to the industry (IFDA's research library and RangeMe's directory), while the other 15 come from adjacent slices and are almost entirely free government data. The table below shows which ones earn a place next to the wholesale share.