Methodology
COMEXT and PRODCOM data granularity
For COMEXT, Eurostat publishes trade data at up to 8-digit granularity (CN-level). Due to business confidentiality concerns, more granular data is only available to Commission staff. The same applies to many PRODCOM datasets. Individual access requests are possible on a case-by-case basis only.
CIF versus FOB
In international trade statistics, the convention is that imports are recorded at CIF (Cost, Insurance and Freight) value. This corresponds to the price of the goods plus the cost of shipping and insuring them to the importing country's border. Exports are recorded at FOB (Free on Board) value. That is the price of the goods loaded onto the vessel at the exporting country's port, excluding international transport and insurance.
Thus, import prices are systematically higher than export prices for the same goods, typically by 5 % depending on the product and distance shipped. Direct comparison between import and export price levels should account for this margin. All price charts on this dashboard are labelled accordingly.
It is not possible to adjust for this reporting bias from COMEXT data alone. The OECD has published an authoritative model along with adjustment tables to correct this effect. However, their adjustment tables stop at 6 digits, while Eurostat data is available up to 8 digits. Therefore, I took the decision not to mislead the user with partly adjusted/partly raw data.
Rotterdam effect
In trade statistics, imports are attributed to the Member State where goods clear customs. But harbours are often just entry points into the EU market. The enterprises or consumers using those goods might be located in another Member State.
Since Rotterdam and Antwerp are the largest harbours in the EU, the Netherlands and Belgium appear as disproportionately large importers of many products, even if many goods are simply in transit to Germany, France, or other member states. Inversely, landlocked countries like Austria or Hungary are under-represented in import statistics. For background, see Eurostat's glossary entry on the Rotterdam effect.
This distortion does not affect graphs where the EU as a whole is the reporting country. However, it affects the reporter-level analysis for external trade, whereby imports or exports with the rest of the world are disaggregated by EU Member State. Within the EU, this distortion necessarily affects partner attribution too.
Finally, this distortion mainly affects bulk commodities transported by sea. It is much weaker for goods produced within the EU and transported by road or rail.
Correcting for this effect is not straightforward:
- COMEXT data does not contain transit declarations, so there is no way to disentangle consignment from final destination based on this single dataset.
-
The
UN Comtrade database, which records country of origin rather than
consignment, could serve as a cross-reference. However
its usefulness is limited:
- Data before and after 2019 is not fully comparable following a methodology revision.
- The most recent years are not yet available.
- The Netherlands and Belgium themselves do not report systematically to the UN their transit trade.
- Ideally, the most rigorous approach would be to use the New Computerised Transit System (NCTS), the EU's digital customs system tracking goods moving under transit procedures. However, this data is not publicly available.
As a result, this dashboard flags graphs where either country features prominently but does not attempt to correct the underlying data.