U.S. Census Foreign Trade — U.S. Trade in Goods by Country (HS 9503/9506)

U.S. Census Foreign Trade — U.S. Trade in Goods by Country (HS 9503/9506) is the federal government's answer to a question politicians argue about and operators plan around: what does America buy from whom, what does it sell back, and what is left over. Every row is a partner-month - exports, imports and balance in millions of USD - and around one hundred partner ledgers carry that shape back through multi-year history. World totals arrive in two forms, seasonally adjusted and not seasonally adjusted; regional rollups span Europe, the Pacific Rim, North America, South and Central America, Africa and the CAFTA-DR and USMCA groupings; and special categories such as Advanced Technology Products sit beside the country tables.

For leisure facilities the entry point is chapter 95 of the harmonized system: 9503 covers toys and games, 9506 covers sports requisites, so the same machinery that produces a national deficit maps where the physical supply of play comes from. See where the note sits among the sixteen cataloged at the leisure facilities data hub, or read the publisher's profile at U.S. Census Bureau. Then Get a sample of this dataset.

Three straight off the ledger's largest single relationship:

Read top to bottom and the quarter argues with itself productively. Imports outran exports in all three months - roughly two-and-a-half times in January, barely under twice by March - yet the deficit narrowed from -12,728.8 to -9,756.0, a swing of 2,972.8 million USD in ninety days. Two properties make those sentences cheap to write. Balance ships as a field rather than something every analyst recomputes, and every figure arrives denominated in millions of USD, so no unit conversion stands between the raw pull and the claim. Because the grain is one row per partner-month, appending March to February is an insert, not a merge project.

Five columns, and each definition was verified against rendered output during the August 2026 research pass - a bar only 1,495 of the 1,744 datasets in Datadory's catalog clear outright.

Country and Month form the spine - the two keys any panel needs before analysis starts. Exports and Imports carry direction, each already in millions of USD. Balance is the subtraction done for you, negative by construction whenever a partner sells more than it buys - which, for consumer goods sourced from Asia, is most months of the year. Five fields is the whole story of the headline table; the depth lives in the layers beneath it.

Additional fields on request: chapter 95 commodity cuts, seasonal-adjustment variants and grouping dimensions stay folded until sample preparation pins exact columns against your pipeline.

• Geography: the United States against roughly one hundred individual partners, plus aggregate views - world totals, continental rollups, the Pacific Rim and treaty groupings including USMCA and CAFTA-DR. One schema serves a single-partner deep dive and a global sourcing map alike.
• Time frame: month-by-month series with multi-year history on each partner ledger, current through 2026 - the January-to-March 2026 China sequence above was verified during research. Monthly grain is what lets a policy headline be checked against the month it claims to describe.
• Granularity: total goods per partner-month in millions of USD, with harmonized-system chapter cuts - 9503 toys, 9506 sports requisites - resolved onto the same spine as a coded layer. Country-first by design; the commodity dimension arrives pinned to your question.

Against the wider catalog this note scores 8/10, above the 7.81 average across 1,744 tracked datasets. Where it ranks: best leisure facilities datasets.

API, files, or your warehouse. Daily, weekly, or hourly.

You choose the channel and the cadence; partner-name normalisation, adjustment-variant labelling and schema stability stay our problem. Rows arrive keyed on partner and month with the balance precomputed, so an incremental delivery appends cleanly and a backfill joins history on the same two keys rather than guesswork. The five-field dictionary travels unchanged across all three channels, chapter-level layers land as extra columns on the same grain, and switching cadence later is a settings conversation rather than a re-integration project. A sample scoped to your partners and months comes first either way.

Six jobs this dataset settles outright:

1. Sourcing-shift detection - toy and sports-equipment import shares migrating between partners flag concentration risk before it becomes disruption; see supply chain mapping.
2. Demand and inventory planning - import arrivals lead the shelf, so a widening partner-month gap reads directly into buying plans; see demand forecasting.
3. Category sizing - measured dollar flows replace survey folklore for how large the import side of play actually is, partner by partner.
4. Exposure screening - deficits by partner and month turn abstract policy headlines into category-specific statements an investment committee can act on.
5. Model training - long monthly panels with stable field semantics are feature engineering done for you; see ML model training.
6. Citation-grade verification - every figure resolves to partner, month and measure; see citation grade research.

Pair the flow side with the price side at BEA PCE Price Index - Recreational Goods & Vehicles and the demand side at BEA Outdoor Recreation Satellite Account.

E-commerce operators rank highest: anticipating sourcing shifts and import-price moves in toys and sports equipment before they reach landed prices is exactly what a partner-month ledger measures - see the e-commerce operators playbook. Journalists, academics and students tie with them, because this is the citation of record for American goods-trade claims - see the journalists, academics and students playbook. Data scientists get a clean monthly panel for demand models - see the data scientists playbook - and market researchers get official flows underneath category sizings - see the market researchers playbook.

Edges worth naming: this is goods trade only, so services, retail shelf pricing and domestic spending live in neighbouring datasets; and the native grain is country-month, not firm, port or SKU. Know the neighbours - Decathlon structured product listings holds the shelf this ledger supplies, and Census Schedule B / HS-Based Trade Statistics works the commodity dimension from the export side.

Three decisions worth making upfront.

First, units and signs: every figure is millions of USD and a negative balance means the partner out-sold the US. Filters written against that convention behave identically across partners; filters that assume positive surpluses quietly drop most of Asia.

Second, pick one adjustment variant and hold it: world totals exist in both seasonally adjusted and not seasonally adjusted forms, and mixing them mid-trend manufactures turns the economy never made. Deliveries label the variant on every row so the choice stays auditable.

Third, scope the layer, not the ocean: a three-partner cut across two years tells you more about usable depth than a shallow pass at everything, and it surfaces immediately whether your question wants the total-goods country table or the chapter 95 coded layer. Sample preparation locks exact columns against your pipeline either way - Get a sample of this dataset.