Is the ban leaking? Chinese EVs after the EU's Oct-2024 countervailing duties

Trade-flow companion to the price wedge (R72). DUAL-SCORE / alternative-track — never folded into any Tier-1 exposure score. The wedge measures scarcity; this desk measures the other half: is the control actually holding, or is the restricted trade re-entering the buyer by another door? Research, not investment advice. Circumvention is INFERRED from the structure of the flows plus public regulatory and corporate records, never asserted as evasion of any particular consignment.

Verdict

On 29 October 2024 the European Commission published Implementing Regulation (EU) 2024/2754, imposing definitive countervailing duties on new battery electric vehicles originating in China, effective 31 October 2024 — 17.0% (BYD), 18.8% (Geely), 35.3% (SAIC), 7.8% (Tesla), 20.7% (other cooperating), 35.3% (non-cooperating), all on top of the 10% MFN duty.

The control has held on its own terms and leaked entirely around them. Chinese BEV shipments into the EU did not surge through a laundered flag. They stayed roughly flat in units and fell by nearly a third in value — while the adjacent, undutied tariff lines at the same, honestly declared origin exploded:

Source: UN Comtrade free public preview API — reporter EU-27 (M49 97), flow imports, annual, HS 8703.80 / 8703.60 / 8703.40, aggregate customs procedure (C00), partner China (156). Duty scope and rates: Implementing Regulation (EU) 2024/2754 and the Commission's own summary of the measure.

Two years after the EU taxed Chinese electric cars, China sold the EU 54% more electrified cars than before the duty. The Commission taxed a powertrain, and the trade moved one HS subheading sideways.

Mode F, and the correction it forces

This is mode F — specification / scope shifting, the mode the steel case opened (Viet Nam's AD duty on Chinese HRC ≤1,880 mm wide → Chinese mills ship 1,880 mm+). Modes A–E all involve something being misrepresented or re-routed: a flag, a factory, a transformation, a port, an end-user declaration. Mode F misrepresents nothing. The controlled party keeps the same origin, the same exporter, the same customer, the same honest customs declaration — and shifts volume into a variant the measure's scope description does not name. There is no relabelling to detect, because there is nothing false to detect.

Either way, mode F defeats the desk's primary tool outright. The volume-implausibility test asks: can this origin plausibly produce what it claims to ship? Here the answer is a straightforward yes — China can obviously build the plug-in hybrids it declares as plug-in hybrids. There is no ~0%-of-world-capacity surger, no namesake shell, no phantom refinery.

But this row corrects mode F's stated corrosive property. The steel case concluded that F is structurally invisible to trade data, because the shifted parameter (plate width) is not an HS distinction — the flow sits inside the same subheading before and after, and detection requires reading trade-remedy scope language rather than customs series. That is one half of the mode, not the whole of it:

> The mode-F visibility rule. Mode F is invisible when the scope boundary cuts > inside a tariff line (steel: width, within one subheading) and legible when the > scope boundary coincides with a tariff line (EVs: powertrain, 8703.80 vs > 8703.60/8703.40). Before concluding a scope-shift is unauditable, check whether > the drafter's scope happened to track an HS split — where it did, the entire > leakage is sitting in the sibling code, fully reported, waiting to be summed.

So the fingerprint is not in the origin column at all; it is in the commodity-code column, and it is visible here only because you read the controlled line and its siblings together.

> The scope-adjacency rule. For any control defined by a product scope narrower > than the strategic dependency it targets, the first place to look for leakage is > not another country — it is the neighbouring tariff line at the same country. > Monitor the control's HS siblings from day one, and read the sum of the family, > not the controlled member.

Read the controlled line alone and the EU measure looks like a policy success: Chinese BEV import value down 31%. Read the family and the same measure looks like a re-mix — total value up 14% and total units up 54%, at an origin whose supply chain still runs on the same Chinese-refined lithium, cobalt, nickel and graphite the anchor action names as its target materials. A plug-in hybrid still carries a battery; it just carries a smaller one, taxed at 10% instead of up to 45.3%.

Controlling for a growing market — the share test

Units alone could be explained by the EU electrified market simply growing. It is not. Against all extra-EU imports on each line, China's share moves in opposite directions across the duty boundary:

Same source and pull as above; denominators are the EU-27 world-total import value on each line.

China lost 12 percentage points of the line the EU taxed and gained 16 points of the line beside it, in the same market, in the same year. That divergence is the signal, and it is not a market-growth artefact: the market grew for everyone, and only China's mix moved across the duty line.

What this case is not — two honest negatives

Turkey is not a laundering leg here. Turkish PHEV-petrol exports into the EU went 275 u (2023) → 19,569 u (2024) → 39,775 u (2025) — a 145× move that would look like a textbook implausible-origin surge to an automated detector. It is not. Ford Otosan's Kocaeli plants are the sole global production centre for Ford's Custom range, the E-Tourneo/Transit Custom PHEV entered production there in early 2024, and Ford Otosan builds ~75% of Ford commercial vehicles sold in Europe. This is a genuine model ramp at a genuine plant, and it is labelled as such — the same treatment the bismuth case gave Korea Zinc. A surge is not a fingerprint until the capacity question is answered.

Part of the hybrid pivot is a global product cycle, not an EU duty response. China's mirror-reported exports to the world show PHEV-petrol rising 138,305 u (2023) → 312,185 u (2024) and non-plug-in hybrids 79,001 u → 261,642 u — a hybrid share of China's electrified exports going 12.3% → 25.8% worldwide, against 8.9% → 19.6% into the EU over the same two years. In 2024 the EU pivot was not faster than the global one. The EU-specific divergence appears in 2025 (EU undutied share 20.4% → 41.2%), for which Comtrade's preview tier does not yet carry Chinese annual data — so the 2025 global comparator is a stated gap, and the duty-attribution rests on the share table above rather than on a world-vs-EU contrast. Both readings are in the same direction; only the magnitude attributable to the duty is uncertain.

The second door: capacity relocation, and the customs-union blind spot

Mode B is running in parallel and is not yet in any import statistic. BYD began pilot passenger-car production at Szeged, Hungary in late January 2026, with series production slated for Q2 2026 — a facility reported at up to €4bn and up to 300,000 units/yr of planned capacity, with the Dolphin Mini as the first model. Ownership is not concealed at any point: BYD builds, in its own name, inside the customs perimeter.

The consequence for this desk is structural, and it is the opposite of every other case in the corpus:

> The customs-union invisibility problem. When the controlled producer relocates > inside the buyer's customs territory, the flow stops being an import at all. > Extra-EU trade statistics do not merely mislabel it — they cease to record it. > A Chinese car built in Szeged and sold in Munich generates no import line, no > origin field, and no duty. The detector's data source disappears, rather than > lying.

Turkey is the adjacent version of the same problem: goods in free circulation move from Turkey into the EU under the EU–Turkey customs union. As of the 2025 data neither leg carries meaningful Chinese volume — Turkish BEV exports to the EU (3,927 u in 2024, 7,531 u in 2025) are of an order consistent with Togg, the Turkish producer, not with Chinese relocation. This is therefore logged as a forward tell, not a finding: the moment Szeged reaches series volume, the correct instrument stops being trade data and becomes registration data by brand.

An open question we are not answering: the duty base

Chinese BEV average declared unit value into the EU fell $23,885 (2023) → $21,631 (2024) → $16,609 (2025), a 30% decline; Korea's comparable BEV unit value fell about 15% over the same span. Because the duty is ad valorem, the declared customs value is also the duty base, and most of these entries are related-party transfers between a Chinese producer and its own European distribution arm — an endogenous base. Mix shift toward cheaper models is a complete and sufficient explanation on the evidence available, and is the more likely one. We flag the structural point and assert nothing: an ad-valorem remedy priced off a related-party transfer value has a base the controlled party partly sets. The Commission's own turn to minimum import prices — guidance issued 12 January 2026 and a first undertaking accepted 10 February 2026 for Volkswagen (Anhui)'s CUPRA Tavascan — is a remedy that only makes sense if declared price behaviour is in question. Testing this properly needs model-level price data we do not have.

Transmission chain

`` Chinese subsidised EV supply chain (CN-refined Li / Co / Ni / graphite) → EU CVD 2024/2754 scopes ONLY "new battery electric vehicles" (HS 8703.80) ├─ leg 1 (LIVE, mode F): same exporters shift mix into HS 8703.60 / 8703.40 │ → +674% units, +427% value, undutied, honestly declared ├─ leg 2 (STARTING, mode B): BYD Szeged (HU) pilot Jan-2026, series Q2-2026 │ → flow leaves the import statistic entirely (customs-union invisibility) └─ leg 3 (POLICY RESPONSE): Commission preparing to extend CVDs to PHEVs (Jun-2026) → scope chase; sibling lines 8703.40 / 8703.70 remain open after it ``

What it implies for the Tier-1 blind spot

A Tier-1 exposure score built on "EU imports of Chinese BEVs" would have recorded improving European exposure to Chinese automotive supply chains from 2024 to 2025. The opposite happened. The upstream dependency the anchor action actually names — Chinese-refined lithium, cobalt, nickel and graphite embodied in the vehicle — rose 14% in value and 54% in units while the headline line fell. The alternative track's job here is to hold the family total next to the Tier-1 line and flag the divergence, exactly as the price wedge holds the ex-China price next to the official one.

The generalisable lesson runs the other way too, and it is the one worth carrying into minerals policy: the EU's forthcoming scope extension to PHEVs will move the trade again, not stop it, because 8703.40 (non-plug-in hybrid, already China's fastest-growing EU line) and 8703.70 sit outside it, and mode B is ramping behind both. Scope-defined remedies against a vertically integrated producer are a chase — the same whack-a-mole the steel case documented (Viet Nam's width duty, then the 1,880–2,300 mm anti-circumvention duty behind it), run one HS subheading wide instead of 420 mm wide. It is also the legal cousin of transformation-washing (mode C): there, a genuine process lawfully confers new origin; here, a genuine engineering choice lawfully confers a new tariff line.

Caveats

• Inference, not proof of wrongdoing. Nothing in this case alleges evasion,

misdeclaration or fraud. Shipping plug-in hybrids instead of BEVs is entirely lawful. That is precisely the point: mode F is legal by construction, which is why it is invisible to enforcement and must be caught analytically.

• Single reporting source for the headline table. All EU-side figures come from

one dataset (Eurostat's declarations, served via UN Comtrade). The 2023–2024 China-side mirror (421,360 u / $12.16bn BEV to 11 major EU destinations in 2023 vs the EU-reported 438,118 u / $10.46bn from all 27) is consistent given the partial destination set and the CIF/FOB gap, but 2025 has no mirror in the preview tier.

• Quantity fields are patchy in this dataset. EU-27 world-total unit counts are

absent for several 2022–2024 line-years, and the UK's 2025 8703.60 unit count (286,972 u against near-flat value) is not credible as reported. Only the China-partner series, which is internally consistent across all four years, is used for unit claims; all share calculations use value.

• HS 8703.40 is a coarse bucket. It contains full hybrids and, in EU practice,

mild-hybrid systems, so its growth mixes drivetrains with very different battery content. The battery-material read-through is therefore directionally right and quantitatively soft.

• Tesla and mix. Part of the BEV value decline reflects non-Chinese brands

shifting EU supply out of Shanghai, not only Chinese-brand behaviour. The share table is the more robust statement; the unit-value discussion is explicitly an open question.

• Data lag. 2025 annual figures are the most recent full year available at the

preview tier as of 1 August 2026; 2026 will re-test both the mode-F leg (after the PHEV scope extension) and the mode-B leg (after Szeged reaches series volume).

Sources

• Implementing Regulation (EU) 2024/2754 — definitive countervailing duties on new

BEVs from China; scope, rates, effective date. Filed as anchor action 2024-10-29-eu-china-ev-countervailing-duties. Commission summary: https://trade.ec.europa.eu/access-to-markets/en/news/eu-commission-imposes-countervailing-duties-imports-battery-electric-vehicles-bevs-china

• UN Comtrade free public preview API (comtradeapi.un.org/public/v1/preview) —

reporter EU-27 (97) and China (156), annual, HS 8703.80 / 8703.70 / 8703.60 / 8703.50 / 8703.40, imports and exports, C00 aggregate customs procedure. All trade figures in this case.

• European Commission, *Commission issues Guidance Document on submission of price

undertaking offers for battery electric vehicles from China*, 12 January 2026.

• European Commission, *Commission accepts price undertaking from Chinese electric

car producer*, 10 February 2026 — Volkswagen (Anhui), CUPRA Tavascan.

• Reporting that the Commission is preparing to extend countervailing duties to

Chinese plug-in hybrids, June 2026 — CnEVPost (https://cnevpost.com/2026/06/20/eu-readies-tariffs-chinese-phevs/) and electrive (https://www.electrive.com/2026/06/19/eu-plans-additional-tariffs-on-chinese-phevs/). Reported measure, not yet a published Official Journal notice; treated as in-progress, not in force.

• BYD Szeged pilot production (late January 2026) and Q2-2026 series target —

electrive, BYD begins passenger car trial production in Hungary, 2 February 2026 (https://www.electrive.com/2026/02/02/byd-begins-trial-production-of-passenger-cars-in-hungary/).

• Ford Otosan Kocaeli as sole global production centre for the Custom range and the

early-2024 E-Tourneo/Transit Custom PHEV launch — Ford Otosan plant disclosures (https://www.fordotosan.com.tr/en/investors/at-a-glance/plants) and Ford of Europe media release, 4 November 2023.