How much of what an American household buys is foreign?

The label on the box answers a different question from the one people mean. A television assembled in Mexico from Korean panels and Taiwanese chips is Mexican to customs and mostly not Mexican in value. This page takes what an average American household actually spends, follows each good back through the trade record to where its value was created, and reports the share that is foreign. The answer is 39.8% of goods spending and 11.3% of the household budget.

Why the trade record cannot answer this on its own

Customs data records imports. It has nothing to say about the shirt sewn in North Carolina or the milk bottled in Wisconsin, because neither ever crossed a border. A trace that starts from the import record therefore credits every dollar of every good to a foreign supplier, and produces a number near 93.4% foreign for a household basket. That is not a dependency measure. It is very nearly one minus the small share that recursion happens to route back home.

So each good is split before it is traced, by the import penetration of the industry that makes it: imports divided by apparent domestic consumption, where apparent consumption is domestic shipments plus imports minus exports. Shipments come from the Census Annual Integrated Economic Survey, imports and exports from Census trade, both 2023. Only the imported part goes through the trade walk. The remainder is domestic production, and it is measured rather than inferred. That remainder is 57.6% of goods value. American value comes to slightly more, 59.5%, because the walk also finds US content inside goods that were imported.

Weights come from the Bureau of Labor Statistics Consumer Expenditure Survey for 2024, all consumer units, mean, where average annual expenditures are $78,535 per consumer unit. Of the 235 survey lines in the basket, 24 are services with no bill of materials and are excluded; the remaining 211 goods lines are 28.4% of the household budget. Country shares below are shares of total goods value: they sum with the 59.5% created in the United States and the 0.70% unresolved residual to one.

Where customs is wrong, and in which direction

Running the same basket one hop deep, stopping where customs stops, gives a different ranking. Both series carry the same penetration split, so the gap between them is the entrepot effect alone and not an artefact of the method.

Mexico (-1.52 pp), Canada (-1.41 pp) and Vietnam (-1.13 pp) are all credited by customs with more than they create: goods shipped from them embody value made elsewhere. China is the one large source that runs the other way (+0.64 pp), because Chinese value reaches American households inside goods that ship from somewhere else. The shipment record understates it.

The basket has two halves

Import penetration is not spread evenly across what a household buys. It is close to bimodal. At the imported end sit computers and computer hardware for nonbusiness use and telephones and accessories, both at 100.0%, with clothing and footwear just behind at 97.7%. At the domestic end sit cigarettes at 3.1%, pet food at 3.4% and cheese at 3.7%. Almost nothing in this basket sits in the middle.

That is why the headline lands near two fifths rather than at either extreme. A household buys a great deal of food, which the United States overwhelmingly makes, and a smaller volume of electronics and apparel, which it overwhelmingly does not.

What this does not cover

The Census survey frame excludes agriculture, so fresh produce, nuts and eggs have no penetration measure. That is 19 of the 211 goods lines and 9.1% of goods weight, and those items keep the assumption of being entirely imported, which makes the foreign share slightly too high. Services are excluded by construction: rent, healthcare and education have no bill of materials, and they are the larger part of a household budget, which is why the budget figure is so much smaller than the goods figure. Comtrade reports country of shipment rather than origin, and recursion mitigates that without eliminating it.

An external check. The San Francisco Fed puts about 11% of US personal consumption expenditures on imported goods and services for 2017 (FRBSF Economic Letter 2019-01). This page gets 11.29% of the household budget from unrelated inputs: Consumer Expenditure weights rather than BEA personal consumption, and a recursive trade walk with Census penetration rather than an input-output table. The two denominators are not identical, so the agreement is corroboration rather than proof.

Computed in github.com/deluair/supply-chain-atlas at commit 1ad68f4 and imported here as three parquet tables. The mapping from each survey line to an HS6 trade code, and from each good to a NAICS industry, is an estimate recorded per item in that repository, not a measurement.