Skip to content
Interface language: English
Oriens / Signal

HS trade data for Korea, Japan and China: how to avoid misleading comparisons

Compare HS trade data with explicit checks on code edition, periods, reporter, partner, units and revisions.

Article language: English

Aeneia concept image: a woman adjusts a gear mechanism beneath two large pale discs.
In this article

Two tables may show similarly named categories but different code editions, periods or units. A comparison can then look precise while pointing the wrong way. With trade data for Korea, Japan and China, first check exactly what each source measures. Calculate shares and changes only afterwards. A calculator is very obedient: it will carefully compute comparisons that should never have been made.

This guide describes a research method. It does not decide the customs classification of a specific product, duty rate or importer obligations. A working category for market analysis needs stated limitations, while classification for a transaction may require separate confirmation. Broader context appears in the Asian market research guide (in Polish).

What question should precede the download?

Write down what you want to compare: imports of a category from the world, the share of a particular partner country or area under the dataset's definition, or a change in flows between two countries. A particular company's share requires another data source. Add the product use and business decision. When ordering 3D printing in China (in Polish), for example, ask whether you are analysing flows of finished parts, material or the provider's service scope. A statistic suitable for choosing a direction may be too broad to assess an individual offer. Without that description it is easy to download a large table whose result does not answer the company's question.

Define geography and period too. In a comparative analysis, countries should share a reference period unless the difference is clearly justified. The latest available month for one country may not be available for the others. Prepare a parameter register before calculating; it saves explaining later why numbers from similarly named sheets diverge.

What is HS and why does its edition matter?

HS is a system for describing and coding goods used in trade classifications. In research, a code provides access to a group of goods. The group's scope can change between editions. The WTO HS Tracker shows code evolution, including splits and mergers, and helps trace transitions between versions.

Record the full category description and edition, not just digits. For a series spanning a classification change, check whether the comparison retains the same scope. If a code was split, simply joining data can create an apparent rise or fall. When uncertain, show a broader aggregate, separate periods or a gap. Explaining a limitation is more useful than a smooth chart based on incompatible categories.

How should a market-research category be selected?

Prepare a product description: function, material, use and significant variants. Compare it with the descriptions of candidate categories. Record why a group is useful for research and what other goods it includes. If the category is broad, treat it as an indicator of the surrounding market. Do not attribute all import value to a product that occupies only part of that group.

Keep a working research category separate from assessment of classification for customs clearance. This article does not replace the procedure for establishing the correct code for an actual transaction. Name the report's scope: “analysis of flows in a selected statistical group”. This supports accurate reading of a market report (in Polish) and discourages later use of the sheet for a purpose it was not prepared for.

What do reporter and partner mean?

The reporter is the country or area whose data you analyse. The partner denotes the other side of the flow under the dataset's definition. Imports reported by Japan from Poland and exports reported by Poland to Japan come from different observations. Do not substitute one without saying so. First establish the reporting side, then use it consistently throughout the comparison.

UN Comtrade's description of data content covers parameters including period, classification, reporter, partner and flow. Recording those settings is part of the result. If you use the other side's data to supplement gaps, label it in the table and conclusion. Changing sources halfway through a series can affect observed dynamics.

Why can imports and exports differ?

Do not expect the two sides of a flow to agree automatically. UN Comtrade's valuation documentation explains recommended CIF bases for imports and FOB for exports. In simplified terms they differ in which delivery costs are included. This is one reason values need not be identical.

Do not remove differences in research with an arbitrary percentage adjustment. Check the definition, registration period and category scope in the particular data. A discrepancy may need further explanation. If it cannot be resolved within the commissioned scope, describe both observations and limit the conclusion. A chart should not suggest more certainty than its underlying material.

How should availability and revisions be checked?

Before analysis open UN Comtrade's data availability information. The documentation explains that publication and updates depend on countries supplying data and datasets can be revised. Record retrieval date and the publication date available with the dataset. For repeatable analysis keep the file used, parameters and transformations performed.

Aeneia conceptual photography

Do not treat an absent row as a confirmed zero. Check whether the entire period, category, quantity unit or only detail is missing. Distinguish unavailable data from actual zero values. If one country has an incomplete year, compare common months or use the latest full period. The choice should also appear in the table title, not be hidden in a footnote.

How should units and calculations be controlled?

Check currency, scale of values, mass and any additional quantity unit. Some files may express values in thousands of currency units, others in whole units. Record the conversion factor and exchange-rate source if you do your own conversion. Do not combine values from different currencies in one column without explaining the conversion.

Value divided by mass is the unit value of the group studied. Do not automatically call it the price of a particular product. A mix of models, qualities and delivery conditions can change the indicator. Before calculating growth, check the comparison base and missing observations. A very small starting value can make a percentage look impressive while contributing little to an entry decision.

When should national sources be consulted?

A national source can help explain detail or presentation. Japan Customs provides a trade-statistics search, and its search instructions describe, among other things, Japanese statistical numbers and aggregates. Additional national subdivisions need checking; do not mechanically compare them with codes of different scope.

Verification is not collecting several similar numbers. Compare definitions, units and dates. If a national dataset and an international database show different values, record the discrepancy and examine possible methodological differences. Then decide which source answers your question. Be consistent throughout the comparison rather than choosing a more favourable value separately for each country.

What does an example without a spurious forecast look like?

Methodological example: a manufacturer of an equipment component wants to compare three destinations. It provisionally chooses a group that also contains other variants. It downloads imports for common full periods, records the HS edition and checks units. The result shows the structure of flows in a broader category. The company does not multiply that value by an invented share to produce a sales plan.

Instead it asks which uses correspond to observed trade, who buys and what role local suppliers play. It then proceeds to market selection (in Polish) and demand verification (in Polish). Statistics identified an area for further work. Only data about the specific segment and conversations with appropriate recipients can bring an offer assessment closer. The Japanese machine-tool industry analysis (in Polish) develops a similar distinction: buying a machine, retrofitting a station and commissioning machining services need different demand evidence.

What should remain in the documentation?

Prepare a dataset card with the full source address, parameters, retrieval date and limitations. Include the original file and calculation sheet. Name the operations: category aggregation, unit conversion, deduplication or selection of periods. Someone taking over the project should be able to reproduce the result without guessing what the author did in the sheet.

If commissioning the analysis, put these materials in the Asian market-research brief (in Polish). For regular updates also decide how revisions to earlier data will be signalled. East Asian market monitoring (in Polish) should distinguish a new period from correction of an old observation. Both can change the conclusion but mean different things to the reader.

How can double counting during aggregation be avoided?

Before summing, check classification levels in the downloaded file. An aggregate row and its detailed components should not enter the same sum. Likewise a flow with the world is not an extra partner to add alongside countries. Record the aggregation rule and check a small table extract to see whether the selected set contains disjoint observations. This check is needed even for official data.

When joining files, establish the row key: reporter, partner, period, classification, code, flow and other parameters used. A repeated key may result from downloading again or omitting an important dimension. Do not remove duplicates solely because values look similar. First establish whether rows actually describe the same observation and record the decision in the transformation documentation.

After calculations, check whether the result makes sense: compare order of magnitude, period coverage and partner completeness. If adding a file sharply changes the result, return to the parameters. A well-prepared table should show which rows entered a sum and why. This allows correction of a technical error without rebuilding the whole methodological conclusion or unnecessarily attributing the discrepancy to market change.

If results from two databases diverge, use the procedure for conflicting sources (in Polish). Compare definitions, period and observation scope first. Choosing the number that better suits a presentation removes the discrepancy only from the slide.

How should the finished table be checked?

  • The business question and product scope are described.
  • The code appears with its description and HS edition.
  • Reporter, partner and flow are unambiguous.
  • Compared periods have consistent scope.
  • Currencies, scales and units have been checked.
  • Missing data is distinguished from confirmed zeroes.
  • Calculations can be reproduced from retained files.
  • The conclusion does not equate imports of a category with demand for the company's own offer.

Aeneia Oriens helps connect trade data with segment analysis and questions for further checking. A well-defined scope supports useful research and clearly names what trade tables alone cannot decide.

Aeneia / Aeneia editorial

Editorial article. Source links are included in the text.

About Aeneia

Brands we have worked with

  • L’Oréal
  • Garnier
  • Netflix
  • Carrefour
  • Warner Bros. Discovery
  • Biedronka
  • HBO Max
  • Wolt
  • Żabka
  • COSRX
  • Embassy of China in Poland
  • Mixa
Aeneia / Initium

From knowledge to a decision

Tell us what you want to put in order. We will agree on where to start.

Contact Aeneia

We partner with and use solutions from:

  • Google
  • OpenAI
  • Microsoft
  • Stripe
  • Anthropic
  • Linear