Is the ban leaking? Russian coal after the EU's Aug-2022 import ban

Trade-flow companion to the price wedge (R72). This case measures whether a control is holding, or whether banned material re-enters the buyer through a laundered origin. DUAL-SCORE / alternative-track — never folded into any Tier-1 exposure score. Research, not investment advice; origin-relabelling is INFERRED from statistical implausibility plus public ownership records, never asserted as smuggling on any single shipment.

Verdict

The EU's coal ban is the cleanest detector-negative in the corpus on the origin line, and a live exposure on the corridor line. Russia supplied 51.8 % of extra-EU coal imports in 2021 and supplies 0.0 % today — 51.28 Mt → 0.00 Mt, not a decline but a stop. No phantom flag carries it back. The volume-implausibility fingerprint that convicts antimony, gallium and tungsten does not appear at all: every origin that gained is a real coal exporter, and the one country the oil-products analogue predicted would be the Trojan — Türkiye — went to 0.30 Mt at peak and back to 0.04 Mt, a route that opened and died.

What replaced Russia is where the exposure moved. Kazakhstan went 0.8 % → 10.5 % of extra-EU coal imports on a flat national export book (28.61 Mt in 2021, 28.54 Mt in 2023) — the tonnes were displaced, not created. And landlocked Kazakhstan reaches the EU by Russian rail through Russian ports. The EU legislature has since said so itself: Art. 5ae(3)(g), inserted by the 18th package (Reg. (EU) 2025/1494), exempts third-country coal "only being loaded in, departing from or transiting through Russia" provided that both the origin and the owner of those goods are non-Russian — the two-layer detector of this whole case series, written into law by the buyer.

This is mode D⁰ — corridor capture / the transit-captive substitute, second instance, and the first on the buyer side of a control. (First: [sulphur](2026-sulphur-russia-export-ban-circumvention-transshipment.md), where Russia's export ban exempted the same Kazakh corridor it could throttle by railway order.) Same corridor, opposite instrument direction, same result: the customs origin table reads fully de-risked while the physical route does not.

1. The buyer line — EU-27 imports of coal (CN 2701), million tonnes

Source: Eurostat Comext DS-045409 ("EU trade since 1988 by HS2-4-6 and CN8"), reporter EU27_2020, flow = imports, annual, QUANTITY_IN_100KG; dataset updated 2026-09-15, extracted 2026-09-18. Extra-EU partners only.

Two readings the table forces:

• The ban held at the border. 2022 is the wind-down year (contracts concluded

before 9-Apr-2022 were executable to 10-Aug-2022, Art. 3j(3)); from 2023 the Russian line is 0.013 Mt, then 0.001, then zero. A bulk commodity cannot hide: coal moves in 60–170 kt cargoes with assayable ash, sulphur and calorific signatures, at ~€100/t. Origin-laundering is a value-density business, and coal has none — the same freight that makes it worth relabelling antimony (€10,000s/t) makes it pointless here.

• Read the denominator. Extra-EU coal imports fell 98.92 → 58.84 Mt (−40 %)

over the same window. Kazakhstan's share more than doubled again in 2025 partly because the book shrank. Share and tonnage are both shown above for that reason.

2. The substitute's own book — Kazakhstan's coal exports (HS 2701, net weight, Mt)

Source: UN Comtrade annual HS, reporter Kazakhstan (398), flow = exports, free public preview API, corridor rows reduced with scripts/py/comtrade_dims.py.

Kazakhstan did not produce more coal for Europe — it re-pointed an existing book. The total is flat (−0.2 %) while EU-declared destinations go from 0.16 Mt to 5.60 Mt (PL+LV+EE) and the Russia line gives up 2.07 Mt. This is the test that separates a genuine substitute from a laundered one in the opposite direction from the usual: a relabelling surger's total exports rise by the banned volume (it is selling somebody else's molecules); a displacement substitute's total stays flat (it is selling its own, to a new address). Kazakhstan is flat.

Two honesty notes on this table. Declared destination is the contracting counterparty, not necessarily the physical discharge port — the 4.21 Mt booked to Switzerland in 2021 is a trader flag whose physical destination this case does not resolve, and it is not counted as EU-bound (the EU's own origin-based figure for 2021 is 0.79 Mt, so those tonnes did not clear EU customs). And the mirror gap is real: Kazakhstan declares 5.60 Mt EU-bound in 2023 where Comext records 6.98 Mt of Kazakh-origin arrivals, a ~20 % discrepancy left labelled, not reconciled.

3. Where the Russian coal actually went

Source: UN Comtrade, importer-side reporters (buyer-side jurisdiction), HS 2701, net weight, Mt.

Türkiye's own coal imports rose 36.53 → 40.08 Mt while its Russian line rose 10.2 Mt: Türkiye absorbed Russian coal, it did not forward it. That is the falsification that matters — the Turkish re-export route is real for refined oil products and absent for coal, and the EU import table above (0.04 Mt in 2025) is the confirmation from the other side of the same trade.

The corridor (why the zero is not the whole answer)

Kazakhstan is landlocked. Its westbound coal moves on Russian railways and loads at Russian Baltic ports, which is why the EU's own port-transaction prohibition collided with its own replacement supply:

• Art. 5ae(1), inserted by Reg. (EU) 2025/395 of 24-Feb-2025 (16th package,

OJ L, 2025/395): "It shall be prohibited to engage in any transaction, directly or indirectly, with ports and locks listed in Part A of Annex XLVII."

• Art. 5ae(3)(g), inserted by Reg. (EU) 2025/1494 of 18-Jul-2025 (18th

package), disapplies that prohibition "to transactions for the purchase, import or transfer of coal falling under CN code 2701 where it originates in a third country and is only being loaded in, departing from or transiting through Russia, provided that both the origin and the owner of those goods are non-Russian."

Read that carve-out as a detector, not as paperwork. The EU wrote a two-limb test — origin AND owner — because a single-limb origin test is exactly what a corridor-captive substitute defeats. It is the first instance in this corpus of a regulator codifying the two-layer method these cases use, and it concedes the residual: EU coal supply now depends on a Russian rail-and-port corridor whose throughput the controlling state sets. Mode D⁰'s map test fires on announcement: when a control's substitute is landlocked behind the controlled state, the diversification is nominal until the route changes.

The ownership layer — Y at industry level, N at shipment level

Kazakhstan's largest coal producer, Bogatyr Komir LLP (Ekibastuz, Pavlodar region), is a 50 % joint venture between Samruk-Energy JSC (the Kazakh state holding) and RUSAL, in place since 2009, per Samruk-Energy's own company page. The same page puts Bogatyr at 70 % of coal mined in the Ekibastuz basin, with design capacity 42 Mt/yr (Bogatyr cut 32 Mt, Severnyi cut 10 Mt), and states that it "supplies coal on market conditions to generating facilities of the Group and third parties located both in Kazakhstan and in Russia."

So the second limb of the EU's own test is not hypothetical: a Kazakh-origin cargo owned by a Russian-part-owned producer fails Art. 5ae(3)(g) on ownership while passing on origin. But this case does not claim Bogatyr tonnes reached the EU. Its stated customers are Kazakh and Russian power plants, and no public shipment record links it to the EU-bound flow. The ownership tell is therefore recorded as Y (industry level) / N (shipment level) — a structural fact about the substitute's industry, not evidence about the cargoes in section 1. An unverified hop is left labelled rather than drawn.

Why it matters for the buyer

• A Tier-1 exposure score reading the EU customs origin line sees **zero Russian

coal* and scores the ban as fully effective. It is fully effective on origin*. The residual sits one layer down: ~10 % of extra-EU coal arrives on rail and ports inside the controlled jurisdiction, from an industry whose largest producer is half-owned by a Russian group.

• **The correct diversification question is not "which flag?" but "which

corridor?" A utility or trader that replaced Russian coal with Kazakh coal moved its country risk and kept its route risk. The sulphur case shows what the throttle looks like when it is pulled: a railway loading order**, not a trade instrument, and reversible in two months.

• The compliance exposure is now ownership diligence on a lawful origin.

Art. 5ae(3)(g) makes the owner of the cargo a condition of legality, which is a KYC question about the seller, not a certificate-of-origin question about the coal.

• Absence of a fingerprint is not absence of exposure. This is the corpus's

cleanest negative on the relabelling detector and one of its larger live exposures. The two findings are not in tension; they are about different layers.

Method & honesty rails

• Inference framing throughout. Nothing here asserts smuggling, and no single

shipment is characterised. The Kazakh flow is treated as genuine Kazakh coal — the evidence supports displacement, not relabelling.

• Sources. EU flows: Eurostat Comext DS-045409 (the EU's own customs data,

buyer-side jurisdiction), queried as https://ec.europa.eu/eurostat/api/comext/dissemination/statistics/1.0/data/DS-045409?format=JSON&reporter=EU27_2020&product=2701&flow=1&time=<year>. Third-country flows: UN Comtrade annual HS free public preview API, corridor rows reduced by scripts/py/comtrade_dims.py (the preview returns one row per partner2 × mode-of-transport × customs-procedure combination; coal moves by sea, rail and road out of Russia, so that dimension is genuinely populated and an un-deduped pull would publish one transport mode as the country total). The Türkiye groups raised the helper's non-replicating-non-summing warning and the aggregate row was used — those two figures are the least precise in this case and are flagged as such. Law: EUR-Lex, Regs. (EU) 2022/576 (OJ L 111, 8.4.2022), 2025/395 and 2025/1494. Ownership: Samruk-Energy JSC company page for Bogatyr Komir LLP. No subscription or analyst sources.

• The originating instrument is not in our register. The coal prohibition

itself — Art. 3j of Reg. 833/2014, inserted by Reg. (EU) 2022/576 of 8-Apr-2022, prohibiting purchase/import/transfer of Annex XXII solid fossil fuels "if they originate in Russia or are exported from Russia", wind-down to 10-Aug-2022 — predates the IPTM register's coverage and is not filed. This case is anchored to the filed 18th-package regulation that carries the corridor derogation. Register-state, not world-state: its absence from the register is a coverage gap, not evidence that no measure exists.

• Dual trigger, noted. Art. 3j binds on origin or export-from, which is

anti-wash wording of the C⁰ family (compare Russian seafood). Whether that drafting or coal's physical economics did the work here is not separable from this data — both point the same way, and the case does not adjudicate between them.

• Unresolved, labelled: the 4.21 Mt Swiss-flagged 2021 Kazakh line; the ~20 %

KZ-EU mirror gap; the physical discharge ports of EU-bound Kazakh cargoes; whether Kyrgyzstan's 0.29 Mt (0.5 % of extra-EU) is own production — it is the only mode-A-shaped candidate in the table and is too small to test.