Read UN Comtrade data without turning a trade gap into a false story

A chart says Country A exported $12 million of a product to Country B. Another says B imported $15 million from A. Before asking where the extra $3 million went, ask whether the two figures describe the same thing.

UN Comtrade can help you examine a claim about trade between countries. A useful comparison starts with the reporting country, direction of trade, product classification, period and valuation. A difference between two totals is a finding to explain; it does not identify a hidden shipment or a particular company.

Read one row as a sentence

Open the official UN Comtrade portal and choose a narrow question before downloading a large table. For example: “What annual value did B report for imports of this product from A?” That sentence is easier to check than “How much business do these countries do?”

The reporter is the country or area submitting the data. The partner is the country or area assigned to the other side of the trade. For a mirror comparison, reverse both the countries and the flow: A’s exports to B versus B’s imports from A. Merely changing the reporter while leaving “exports” selected asks a different question.

Copy the row’s labels with its number. A saved number without its unit, period and product definition is difficult to interpret later. Keep the downloaded file unchanged and make a separate working copy for filtering. The site’s CSV import guide explains how to preserve code columns that a spreadsheet might otherwise reformat.

Make a comparison ticket before calculating

Use three columns: the first record, its proposed mirror, and the unresolved difference. Fill every row below. An empty cell means the comparison is not ready, rather than permission to assume a match.

This ticket is a working method, not an official UN form. Its purpose is to make the first wrong assumption visible before it becomes a percentage in a headline.

The $3 million difference: a fictional exercise

The countries and amounts in this exercise are invented. You have an annual export record for A→B worth $12 million and an annual import record for B←A worth $15 million. You initially write “B received 25% more than A sent.” The arithmetic, (15−12)/12, is correct. The sentence is not yet supported: value received and physical goods sent are different measures.

First check that the records use the same product definition and period. Next read valuation metadata. UN guidance recommends CIF-type values for imports and FOB-type values for exports. Import valuation may therefore include international freight and insurance that the export valuation does not. National practices need checking; do not subtract a guessed shipping percentage to force the figures to agree.

Now imagine your metadata note establishes this valuation difference, but gives no amount for transport costs. Your conclusion can say that the reported values differ by $3 million and are on different valuation bases. It cannot allocate the whole gap to freight. That unresolved amount belongs in the next research question, not in an accusation.

If the original claim concerns tonnes, start again with comparable weight fields. A value comparison cannot settle the physical amount by itself. If one side has no usable quantity, record that limitation rather than borrowing a price from another product or year.

Why the mirror may remain uneven

The UN’s bilateral-asymmetry guidance describes several reasons exports and partner imports can diverge, including valuation, timing and partner attribution. A departure near year-end can be recorded in a different period at the other end. That is a possible explanation to test with metadata, not a fact established by the size of the gap.

Partner labels deserve their own check. A product’s origin and the country from which it was dispatched can differ. Comtrade has a second-partner dimension for consignment. Record the chosen filters explicitly; “all origins dispatched from A” is not the same population as “goods originating in A, dispatched from anywhere.”

Also resist “repairing” an annual total by summing monthly downloads. The UN explains that annual and monthly series can differ because of separate processing, revisions, confidentiality treatment and classification changes. Preserve both results and their definitions. A spreadsheet sum is not automatically a more authoritative annual figure.

If you compare product categories over several years, check whether the series uses reported or converted classifications. Read the conversion notes before treating an apparent category jump as a change in the underlying trade. A code is a label within a particular classification, not a timeless product identity.

Keep the country finding separate from the company claim

A supplier may point to a national trade chart as evidence of its own exports. That changes the question. UN Comtrade does not provide buyer/supplier identities or individual shipment dates. A country-product total cannot establish that a named firm supplied the goods.

For that claim, write two separate findings: what the national series reports, and what evidence connects the firm to the claimed activity. Follow the company-claim verification workflow for the second task. If several public records need reviewing together, OSINT Jet’s manual investigation request is a place to scope that question. Describe the missing connection; do not request a guaranteed buyer list from an aggregate dataset.

Before handing over the research, include the two downloaded records, your comparison ticket and a short conclusion: “These reported values differ on the following stated bases. These differences remain unexplained.” Someone else should be able to rerun the selection without reconstructing your assumptions from a chart.