September 2026: the U.S. trade gap narrowed.
The goods story did not.
Down $0.7B from revised August
September’s U.S. trade deficit narrowed to $105 billion, but real goods trade, computer imports and gold accounting show a more complicated picture.
The Census Bureau and Bureau of Economic Analysis released the figures on November 4, 2026. The report covers September, not the week of publication. Its headline values are seasonally adjusted but expressed in current dollars. That combination matters: predictable calendar effects have been addressed, while changes in prices remain embedded in the totals.
Exports reached $321.5 billion, up $5.8 billion from August. Imports reached $426.6 billion, up $5.2 billion. The deficit declined $0.7 billion from a revised $105.7 billion in August. The separately rounded figures do not reproduce every reported change to the last tenth of a billion, so subtracting rounded headline numbers is not a reason to overwrite the agencies’ published balance.
A useful first distinction is between the size of trade and its balance. A deficit is the difference between two flows. It can shrink when both flows rise, as happened here, or when both fall. Those situations have different implications for businesses even if the subtraction yields the same result. September therefore was not a simple story of Americans buying less from abroad.
The small monthly improvement sits beside a larger recent gap
September’s monthly goods-and-services deficit was lower than August’s, but the three-month average rose $11.3 billion to $101.2 billion. Average exports increased $1.4 billion to $316.0 billion, while average imports increased $12.7 billion to $417.2 billion. That moving window gives a different signal from the latest month alone.
Compared with the three months ending in September 2025, the average deficit was $36.5 billion larger. Average exports were $28.9 billion higher, but average imports were $65.4 billion higher. The recent trading period therefore showed a much wider gap than the equivalent period a year earlier, even though September improved slightly from August.
Meanwhile, the year-to-date deficit was $92.4 billion, or 12.2 percent, smaller than in the first nine months of 2025. Exports were up $297.3 billion, or 11.6 percent, and imports were up $204.9 billion, or 6.2 percent. That is not a contradiction. A January-through-September comparison includes months that are absent from a July-through-September average. A claim about improvement over the year and a claim about deterioration in the recent three-month window can both be accurate.
Computers moved the import total, but the category is not a diagnosis
On a Census basis, goods imports increased $8.0 billion in September. Capital goods rose $11.6 billion, a larger gain than the total because other categories moved in the opposite direction. Computers increased $7.6 billion and computer accessories increased $5.7 billion, while semiconductors decreased $1.2 billion.
These are product classifications, not a survey of purchasers’ intentions. The figures establish a large rise in the value of computer-related goods crossing the border. They do not, by themselves, establish how much went into artificial-intelligence facilities, which company made a particular purchase, or when imported equipment began productive operation. An article or investment argument that attaches those conclusions to the whole category needs evidence beyond this release.
Nor should the fall in semiconductor imports be read as proof that computer demand weakened. A finished computer and an individual semiconductor occupy different product classifications. Changes in the mix of complete systems, parts and components can pull those lines in different directions. This report does not identify the exact assembly decisions behind the month’s movement.
Consumer-goods imports rose $2.4 billion, including a $2.9 billion increase in pharmaceutical preparations. The pharmaceutical gain exceeded the category’s net gain, showing that other consumer-goods movements offset part of it. Industrial supplies and materials fell $3.7 billion, including declines of $2.7 billion in nonmonetary gold and $1.3 billion in crude oil. Category totals are the net result of multiple products, not a list in which every highlighted increase adds to the headline without an offset.
That is why a container photograph cannot explain the entire release. The value inside one container can differ sharply from the value inside another, and some traded goods move by air, pipeline or land. The trade report measures dollar values under defined statistical rules; it is not a direct measure of container counts, port congestion or the number of trucks passing a checkpoint.
Exports rose in more than one place
Goods exports on a balance-of-payments basis increased $5.1 billion to $211.0 billion. On a Census basis, the increase was $5.0 billion. Capital goods added $3.8 billion, with computers up $1.3 billion, computer accessories up $0.8 billion and civilian aircraft up $0.7 billion.
Foods, feeds and beverages increased $1.9 billion, including a $1.3 billion rise in soybeans. Consumer goods increased $0.8 billion, including $0.7 billion in pharmaceutical preparations. Those figures indicate that September’s export increase was not confined to one type of product.
The agricultural line still requires restraint. The published amounts are seasonally adjusted dollar values. They are not a direct count of bushels, a measure of farm profitability, or evidence that every producing region received higher prices. Quantity, price, destination and timing can matter differently to a farmer, an exporter and a shipping company. A larger export value alone does not settle those questions.
Industrial supplies and materials exports declined $1.0 billion. Within that category, nonmonetary gold fell $3.4 billion while crude oil rose $1.2 billion. The opposite directions again demonstrate why interpreting the broad category as a single commodity market would be misleading. A reader trying to understand energy demand should not substitute the whole industrial-supplies line for the petroleum detail.
Civilian aircraft warrant particular care when following monthly trends. The release’s methodological notes explicitly identify aircraft as highly variable and suggest analyzing that trade separately when studying trends. A valuable aircraft delivery can alter a monthly total without demonstrating an equally sudden change in the entire manufacturing economy.
Two goods deficits can move in opposite directions
The headline goods deficit, on a balance-of-payments basis, decreased $0.7 billion to $135.9 billion. Yet the Census-basis nominal goods deficit increased 2.2 percent, and the real goods deficit increased $3.5 billion, or 3.0 percent, to $119.5 billion in chained 2017 dollars. Those are different statistical measures, not competing claims that must be averaged together.
Census goods data primarily describe merchandise moving across the customs frontier, using information collected by Customs and Border Protection. BEA adjusts those data to align them with the concepts used in the international economic accounts. The adjustments address coverage, duplicate recording and valuation, among other issues. A balance-of-payments series therefore should not be interchanged casually with a Census-basis series.
In September, goods imports increased $8.0 billion on a Census basis but only $4.4 billion on a balance-of-payments basis. Net import adjustments decreased $3.6 billion. The release specifically attributes the difference in direction between the two nominal goods deficits to a balance-of-payments adjustment for nonmonetary gold imports.
This is not evidence that one agency counted ships incorrectly. It is a warning that the same word, “imports,” can refer to a total assembled under different accounting boundaries. Before comparing figures, the reader needs to check the basis, whether services are included, the seasonal treatment and whether the dollars have been adjusted for prices.
The real goods figures address a separate question: how trade changed after adjustment for price movements. Real goods exports increased just $0.2 billion, or 0.1 percent, to $152.2 billion. Real goods imports increased $3.6 billion, or 1.4 percent, to $271.6 billion. The comparable nominal Census-basis increases were 2.5 percent for exports and 2.4 percent for imports.
The sharper rise in nominal exports than in real exports means the dollar increase should not be described as an equivalent increase in physical output. The real series uses a chain-weighted method and 2017 reference dollars; it is not a literal sum of today’s shipments priced using a single old price list. It is useful for separating price-related changes from the current-dollar story, but it still has its own definitions and revision procedures.
Why gold needs a separate explanation
“Nonmonetary gold” does not simply mean jewelry, and it does not mean all official reserve gold. BEA’s guidance distinguishes monetary gold held by monetary authorities as a reserve asset from other gold. The balance-of-payments nonmonetary category can include bullion, bars, powder, certain coins and other stages of manufacture. Gold jewelry is recorded separately in consumer goods.
The Census nonmonetary-gold category and the balance-of-payments category do not have identical coverage. BEA combines relevant Census categories and makes an adjustment for certain transactions between foreign official agencies and private U.S. residents that customs documents do not capture. For example, a change in ownership of gold held at the Federal Reserve Bank of New York can be recorded as an international transaction even without a shipment across the customs frontier.
The September release does not provide a transaction-by-transaction explanation of the month’s gold adjustment. It would therefore be wrong to invent a particular central-bank sale or trading strategy as its cause. The supported conclusion is narrower: the published adjustment was large enough to change the comparison between the Census-basis and balance-of-payments goods deficits.
Gold is treated differently again when BEA measures domestic production. Much nonmonetary gold trade reflects purchases as a store of wealth rather than an input into current production. For the national economic accounts, BEA removes the international-account gold export and import amounts and replaces them with an adjustment based on domestic gold production minus industrial use.
That distinction prevents a tempting shortcut: taking the latest trade headline and translating it directly into a GDP contribution. September’s goods-and-services balance is an international-trade measure. GDP incorporates its own definitions, price adjustments and treatment of gold. A precise growth contribution must come from the national accounts rather than from an improvised subtraction of monthly trade headlines.
Services offset part of the goods gap
Services exports increased $0.7 billion to $110.5 billion, while services imports increased $0.8 billion to $79.7 billion. The services surplus decreased by less than $0.1 billion to $30.9 billion. It remained a substantial offset to the $135.9 billion goods deficit.
Travel exports increased $0.3 billion and charges for the use of intellectual property increased $0.1 billion. On the import side, travel increased $0.5 billion and transport increased $0.2 billion. The report combines tangible merchandise with services that may not involve a box crossing a port at all.
Travel is especially easy to misread in everyday language. The relevant distinction in international accounts is the transaction between residents and nonresidents, not simply whether an American sees an airplane or a hotel. The release’s explanatory notes describe services categories separately from merchandise documentation. A goods-only trade chart cannot represent the whole relationship between two economies.
There is also a rounding issue here. Subtracting the displayed $79.7 billion from $110.5 billion produces $30.8 billion, while the official services surplus is $30.9 billion. Published components are rounded individually. Use the published balance rather than labeling that tenth-of-a-billion difference an unexplained discrepancy.
Country comparisons need the same statistical basis
The seasonally adjusted Census-basis goods deficit with Taiwan increased $3.6 billion to $21.9 billion. The European Union deficit increased $3.2 billion to $14.2 billion, while the Canada deficit decreased $2.7 billion to $4.4 billion. These figures locate changes in merchandise trade; they are not balances for goods and services combined, or calculations of each country's value added throughout a supply chain.
The release cautions that seasonally adjusted country totals do not sum to seasonally adjusted commodity totals, because their aggregations and seasonal models differ. For a company sourcing components, those data can indicate where to investigate further. They cannot replace supplier-level delivery, origin and contract information.
Imports are not automatically a subtraction from economic well-being
The familiar GDP formula subtracts imports to avoid counting foreign production as U.S. output. BEA explains that spending on consumption, investment and government purchases can contain imported products. Subtracting imports removes that foreign portion from the measurement of domestic production; it does not, by itself, describe the purchase as an economic loss.
Consider a business acquiring imported equipment. The expenditure can appear in investment while the imported portion is removed through the trade adjustment. That accounting prevents the imported equipment from being counted as newly produced in the United States. It does not measure the equipment’s future productivity, the return on the investment or the jobs associated with its installation and use.
Tariffs require another distinction. BEA’s national-accounts guidance says the value of goods imports does not include customs duties. Duties accrue to the federal government and are reflected in taxes on production and imports. A change in the merchandise import value is therefore not a complete reading of the importer’s landed cost, and it is not the government’s tariff-revenue statement.
This report also cannot isolate the effect of any single trade policy. Exchange rates, demand, shipment timing, product prices, inventory decisions and accounting adjustments can all matter. Demonstrating a policy effect requires a comparison designed to distinguish those influences, not simply pointing to a deficit that moved in a preferred direction for one month.
What to carry into the next release
The strongest reading of September keeps four observations together. Both trade flows expanded. The overall current-dollar deficit narrowed slightly. The three-month average deficit widened. And price-adjusted Census-basis goods trade showed a larger deficit, with gold accounting explaining an important difference between nominal measures.
The next scheduled release is December 8, covering October 2026. Readers should first check revisions to September before interpreting the new month. The current release already revised August goods and services exports upward and revised both goods and services imports upward. Those updates are part of the publication process, not an exception that can be ignored whenever they complicate a comparison.
The real series also receives revisions incorporating updated price indexes. Goods records are administrative data, but that does not make them free of error: the release describes reporting, timing, capture and low-value estimation issues. For detailed commodity claims, a small movement deserves more scrutiny than a sweeping conclusion based on the absence of a survey margin of error.
Reporting and sources
Newjinmedia prepared this analysis from the November 4, 2026 Census Bureau and BEA trade release and BEA’s published methodological explanations. It is document-based reporting, not firsthand reporting from a port or a trading firm. Archival photographs illustrate trade infrastructure and products; they do not depict the specific September transactions in the report.
Comments
Post a Comment