Method and limits

← CBAM Exposure

Every number in this section is an administrative default, marked up on purpose. A real importer with supplier data will usually pay less, and every importer's obligation is further reduced by two deductions this section does not model.

The annex and its correction

The values are Annex I of Implementing Regulation (EU) 2025/2621 as corrected by Implementing Regulation (EU) 2026/1740, which replaced Annexes I and IV in full on 3 August 2026 and applies retroactively from 1 January. The numbers barely moved (66 of 10,503 comparable rows, mostly down by a percent or two), but the table changed shape, and most of the original annex's defects were fixed at source. The first annual declaration, covering 2026 imports, is due in 2027.

The mark-up is set per product group

The defaults carry a mark-up of 10% in 2026, 20% in 2027 and 30% from 2028, with fertilisers at 1% in all three years. A flat 10/20/30% would overstate every one of the 2,457 fertiliser rows by nine points in 2026 and twenty-seven by 2028.

The price is a parameter

There is no clean free public API for EU ETS spot, so the carbon price is an assumption, stated on the overview and applied to the regulation's tonnage. The range charted there, €60 to €120 a tonne, is roughly where EUAs have traded since 2022, with room above it.

What this is not

• The figures are gross. What an importer surrenders is reduced by two deductions that nothing here models: one for the EU ETS allowances EU producers of the good still receive free (a benchmark per good, times a factor that phases out from 2026 to 2034), and one for any carbon price effectively paid where the good was made, under Article 9 of Regulation (EU) 2023/956. When this was written the benchmarks were provisional and the Article 9 rules not yet adopted, so neither can be applied yet with confidence.
• A CN code alone does not always identify a row. In the original annex, 2523 10 00 was both white clinker and grey clinker, whose default values differ by more than a factor of two. The correction gives those two 10-digit TARIC codes, 2523 10 00 10 and 2523 10 00 90, so that trap is closed, but the annex still prints 4- and 6-digit headings above the rows that carry the numbers, and classification to the right description remains the importer's problem.
• The annex's defects are reproduced rather than corrected. The seed transcribes the regulation as it stands. A legal instrument is not this project's to tidy up, and the correction is the argument for that: the original annex's quirks were fixed by the body that wrote it.
• The mark-up is asserted, not read off the annex. The corrected annex publishes only direct, indirect and total, so the phase-in comes from a seed stating what the articles say. Same numbers, weaker provenance.
• The grid factor shown elsewhere in this warehouse is not the annex's. The regulation's own electricity emission factors come from IEA data under a non-commercial licence, which this project deliberately does not redistribute. The indirect part of the annex cannot be reconciled against the OWID-derived factors on the Scope 2 page. They sit beside each other; they are not the same measurement.

The underlying table is marts.fct_cbam_exposure, and the transcribed annex is the cbam_default_values and cbam_goods seeds, with the phase-in rates in cbam_markup_schedule. The mart ships in the data release both as Parquet and inside the DuckDB file; the seeds are in the DuckDB file's main schema only.