Did China's export controls bite?

Source: Eurostat Comext, monthly detailed trade (CN8). Computed values: out/export_controls.json.

1. The six controls

Each was checked against the announcement itself, in the Chinese original. The October 2025 expansions were suspended on 7 November 2025 and are not tested. The import data end in Jul 2026.

Source: the Ministry of Commerce announcements listed under References. The customs codes for each line are in the filing.

2. The test, as filed

NO CONTROL SHOWN TO BITE, AS FILED

Each controlled good was set against a comparison good that China also dominates and that no control we found covers — unwrought magnesium for the metals, talc and baryte for graphite, ferrite magnets for rare-earth magnets, lanthanum and cerium compounds for the heavy rare earths. These are coarse comparisons: each has its own market and can move for its own reasons (magnesium's price swung widely in 2021–2022), and a comparison bought almost entirely from China, like magnesium, also nets out anything that hit all Chinese exports at once, which pushes towards finding no effect. Ferrite magnets are a substitute for rare-earth magnets, so a switch between them moves both sides. The monthly gap between the two was compared before and after, with an anticipation window between announcement and entry into force, errors robust to autocorrelation^[14], and a correction for running six tests at once (Holm’s). The filing read months 7–12 after entry into force. A control bit if imports from China fell by at least 30% against the comparison and either the total fell by 20% or the price rose by 20%. And the filing said to check first whether the design could have seen a 30% fall at all: if not, the control is untestable, not a null.

Five of the six EU controls are untestable by that rule, and rare-earth magnets show no visible fall. The quantities are thin and jumpy month to month; only the magnet series is steady enough to see a 30% fall.

Source: Eurostat Comext, monthly detailed trade (CN8). Computed values: out/export_controls.json.

Quantities are differences of the inverse hyperbolic sine of kilograms, which keeps the months when nothing arrived from China; such a difference is not a percentage, and when a series approaches zero it is not a log change either^[15][16]. The unit value is a true log difference and its percentage is shown. The last column reverses the order of the filed rule so that an interval already beyond the threshold decides the reading whatever the power; that correction was made after the estimates were seen, and it is reported beside the filed reading, not in place of it. It also reads the quantity gaps against a 30% line drawn for logs, which is least valid for antimony, whose China series has months with almost nothing arriving.

3. What the series show

Leave the comparison aside and look at the tonnes and the prices themselves: the mean month in the two years before each announcement against the mean month in months 7–12 after the control.

Source: Eurostat Comext, monthly detailed trade (CN8). Computed values: out/export_controls.json.

These are before-and-after means, not effects. Buyers can stock up before a control and draw down afterwards, and a customs unit value moves with the mix of grades shipped as well as with the price.

Source: Eurostat Comext, monthly detailed trade (CN8). Computed values: out/export_controls.json.

Antimony: imports from China all but stopped; total imports did not fall

EXPLORATORY

EU imports of antimony from China fell from 190 tonnes a month to 8.4, while antimony from all origins went from 1,664 to 1,719 tonnes a month: imports from other origins made up the difference. China supplied only about 11% of EU antimony tonnes even before, so the price below is mostly that of antimony from elsewhere. The unit value of all EU antimony imports rose from €11.4 to €47.0 a kilogram, 4.1-fold, and not because the mix changed: unwrought antimony, most of the tonnage, rose about as much on its own, and ores and concentrates rose too. The estimate against magnesium (×6.3) is larger because magnesium itself got cheaper over the same months (×0.65).

Much of this began before the control. Imports from China averaged 221 tonnes a month at a unit value of €11 over most of the pre-period, but only 36 tonnes a month at €13 in the four months before the announcement (Apr 2024 to Jul 2024). The squeeze was under way before August 2024, and the announcement formalised it; this design cannot separate the two, nor the EU price from China's December 2024 ban on antimony exports to the United States, which fell inside the same window. If there was a bite in Europe, it shows in imports from China and in the price, not in total import volume.

Bismuth: less from China and less in total, at a higher unit value

EXPLORATORY

Bismuth from China fell from 132 to 56.9 tonnes a month in months 7–12, and total imports from 145 to 102. Bismuth from other origins rose, from about 13 to 45 tonnes a month, and filled part of the gap, not all of it. The unit value of all bismuth imports rose 2.6-fold (×3.7 against magnesium). Of the six, bismuth is the only one where all three outcomes — imports from China, total imports and the unit value — passed the filed thresholds against its comparison, each with its interval excluding zero. The squeeze eased late in the window: imports from China averaged 16 tonnes a month at €36 a kilogram from Jun 2025 to Oct 2025, and 73 tonnes at €24 from Nov 2025 to Feb 2026.

The rest

Gallium and germanium are too thin to test: a few tonnes a month, with the unit value up 1.4-fold (+117% against magnesium, more than half of which is magnesium getting cheaper). The customs lines cover the unwrought metals and powders, not the oxides and other compounds the control also covers. Natural graphite from China rose in tonnes, and as filed the reading is untestable; the synthetic and spherical graphite the control also covers is not in the customs lines the filing locked, so this says nothing about them. The 2023 announcement also revised controls on graphite in place since 2006, so the years before it are not an uncontrolled baseline. Rare-earth magnet imports from China rose relative to ferrite magnets, which rules out a 30% fall against ferrite, not any effect; the customs line holds every neodymium, praseodymium, dysprosium and samarium magnet, while the control covers only samarium-cobalt and terbium- or dysprosium-containing magnets, so an effect on those would be diluted. The heavy rare earths — gadolinium, terbium, dysprosium metals and compounds — halved in total, but the China-origin series is a few tonnes a month and the unit value jumps by factors of ten from month to month, so neither says whose supply moved.

4. Who replaced China?

A follow-up asked where the replacement came from: from countries that produce the metal, or from countries that might be passing Chinese material on. It was chosen after the results above were known, and filed as an amendment before any origin-level figure was read. The origins that together supplied at least 80% of the gross increase were each classed as a producer if their own mine or processing output, in the British Geological Survey's world statistics, covers at least a year of what they now ship to the EU, and otherwise a non-producer. A non-producer whose shipments at least tripled, by 10 tonnes a month or more, became a transit candidate, and its own imports from China were then checked in the UN Comtrade records it reports.

Source: Eurostat Comext, monthly detailed trade (CN8) · BGS World Mineral Statistics. Computed values: out/export_controls_origins.json.

Antimony: re-sourced, mostly from South-East Asia. The named rises are gross gains, not a one-for-one replacement of Chinese metal: other origins added 392 tonnes a month against the 182 China lost, while Tajikistan (-77 tonnes a month), Oman (-44 tonnes a month), Türkiye (-26 tonnes a month) shipped less. Myanmar, which mines antimony, is the largest gainer and the only one whose mine output covers its later EU shipments: 35% of the gross gain. Viet Nam and Thailand fail the mine screen; the screen cannot say whether they refine ore mined elsewhere or re-export. Malaysia, which had shipped none before, is the one transit candidate (67 tonnes a month after). By the filed check it is not consistent with rerouting: the unwrought antimony it reports importing from China was 0.14 tonnes a month before and 0.03 in months 0–12 after (a month with no record counts as none). That rules out a large rise in direct imports of Chinese metal on that line, not stocks, processing elsewhere or other customs lines. A side look at three months, not a test, found Malaysia importing antimony ore from Myanmar, Türkiye, Thailand and a partner recorded as the United Kingdom.

Source: Eurostat Comext, monthly detailed trade (CN8) · BGS World Mineral Statistics. Computed values: out/export_controls_origins.json.

Bismuth: South Korea, largely redirected from American buyers. Korea's shipments to the EU rose from 1.5 to 26 tonnes a month, which made it a transit candidate, since it mines no bismuth. By the filed check it is not consistent with rerouting: the bismuth it reports importing from China fell, from 7.8 to 2.6 tonnes a month (11 of 13 months reported), and was too small even before to cover the extra EU sales. A check added after review, on Korea's own export records, shows where the metal came from: Korea's bismuth exports to the world rose modestly, from 68 to 81 tonnes a month, while its exports to the United States fell from 61 to 33 and those to EU countries rose from 0 to 26 (February to December 2025, the 11 months Korea has reported). The fall in shipments to the United States is about the size of the rise to Europe. Korea refines bismuth as a by-product of zinc smelting — about 1,000 tonnes a year, among the largest outside China^[13][11] — so that metal is Korean output; more of the change was in where it was sold than in how much was made.

Source: Eurostat Comext, monthly detailed trade (CN8) · UN Comtrade, monthly, as reported by the importer. Computed values: out/export_controls_origins.json.

Neither check is proof: material can be processed in a third country, declared under another customs line, or come from stocks. What the records show is that neither candidate's reported imports from China rose on the line checked.

5. Why these two?

One simple prior is that controls move trade most where China holds the largest share of world output, where there are fewest other producers, or where Europe relied on China most. These are crude proxies, not a supply-chain model. The same amendment set the question before reading the production figures and said the answer would be reported whichever way it fell; with six rows it is a description, not a test.

Source: Eurostat Comext, monthly detailed trade (CN8) · BGS World Mineral Statistics. Computed values: out/export_controls_origins.json.

Mine output for antimony, bismuth and graphite; processed output for gallium, germanium and rare earths (both rare-earth rows use the same rare-earth figure, not magnet making). USGS gives mine output for antimony^[12] and refinery output for bismuth^[13]. *The British statistics record China's bismuth output as 1,804 tonnes in 2019–2021 and exactly 1,800 in 2022–2024, a flat figure that looks like a placeholder; the other two sources put China at three-quarters or more.

The two that moved sit at opposite ends. Antimony has the lowest Chinese share of the six on both the filed BGS figures (38%) and World Mining Data (40%), and the USGS summary's 59% is still below every other row. Bismuth looks lowest only on the flat BGS figure; on World Mining Data (75%) it is third of six, and on USGS refinery output (82%) it is high, though still below gallium and germanium. On Europe's own reliance they also split: 11% for antimony, 91% for bismuth. No single column picks out these two, and gallium and germanium, which the simple prior would pick first, were too thin to test. The table describes six rows; it does not reject the idea that controls bite where China's position is strongest, especially at the refining stage the controls actually cover, which the open data measure for bismuth but not for antimony.

6. The United States: no answer

The same test was filed for US imports, and the comparisons filed for it do not support a reading. US imports of magnesium from China fell by about 90% between 2021 and 2025, so any good compared with it looks as if its China imports rose; US talc imports jumped six-fold in recorded kilograms in 2023 with their value flat, which looks like a change in recording; and US ferrite magnets are counted in pieces, not kilograms. Magnesium was found after the first run and the other two after the second, each logged in the filing. There is no clean US customs line for the heavy rare earths, and US general imports include goods later re-exported. No replacement comparisons were chosen after seeing the data, so the US half has no reading, including the outright ban on gallium, germanium and antimony to the United States (C3)^[4].

Source: filing, deviations 3 and 7; US Census Bureau, international trade API (HS10).

What to watch

Two suspensions end in November 2026. China's October 2025 announcements — among them five more medium and heavy rare earths (holmium, erbium, thulium, europium, ytterbium)^[9] and lithium batteries and artificial-graphite anode material^[10] — are suspended until 10 November 2026^[8], and the ban in principle on gallium, germanium and antimony to the United States until 27 November 2026^[5]. They cover goods this page does not follow: other rare earths, battery anodes, and a ban on shipments to the United States that EU records do not see. What carries over is the method: in the customs lines those goods fall under, a fall in tonnes from China, a rise in the unit value of all imports, and new origins whose own records should be checked. Eurostat's monthly figures run about two months behind (July 2026 was the latest on 21 September), so a change in November 2026 would show in early 2027.

What the record can and cannot tell a buyer. In the two cases where something moved, the unit value of all EU imports rose (four-fold for antimony, 2.6-fold for bismuth) more clearly than total tonnes fell: antimony tonnes held, bismuth tonnes fell by about a third. These are customs averages, not contract prices, and not clean effects of the licences. Replacement came from countries whose own mines do not cover what they ship, and in bismuth's case largely from metal redirected away from other buyers, so a mine-based view of supply would have missed where it came from. And monthly public customs data lag and are noisy: they are a warning signal alongside supplier licences, inventories and contract prices, not a substitute for them.

What this can and cannot say

It is the EU's import record, to Jul 2026, against coarse comparison goods, not a causal estimate. Magnesium is bought almost entirely from China, and its EU unit value roughly tripled between early 2021 and mid-2022 before falling back; the years the test uses as “before” contain parts of that swing, so every price estimate against it carries some of it, which is why the raw multiples are shown beside it. Customs value is not a contract price, and a unit value of a small-volume good can move with the mix of what was shipped. Material shipped ahead of a control shows up as anticipation, licences granted are not observed, and routing through a third country is inferred from the pattern, not seen.

What it does say: in Europe's customs record, antimony and bismuth show large movements around their controls — imports from China fell, antimony's to almost nothing, and the unit value of all imports rose several-fold — while the filed test, built to guard against reading noise as effects, could not certify either as a bite. For antimony the fall began before the control, so the record cannot say how much of it the control caused. Both readings are reported because a pre-registered test that cannot see what the raw series show is a finding about the test as much as about the controls.

References

1. Ministry of Commerce and General Administration of Customs of China (2023). Announcement 2023 No. 23 on export controls on gallium- and germanium-related items (in Chinese). www.mofcom.gov.cn/zcfb/dwmygl/art/2023/art_52b9a321087f402bb3d310d18b07967e.html
2. Ministry of Commerce and General Administration of Customs of China (2023). Announcement 2023 No. 39 on optimising the temporary export controls on graphite items (in Chinese; text as republished on the Ministry’s export-control information site). exportcontrol.mofcom.gov.cn/article/zcfg/gnzcfg/zcfggzqd/202310/912.html
3. Ministry of Commerce and General Administration of Customs of China (2024). Announcement 2024 No. 33 on export controls on antimony and other items (in Chinese). www.mofcom.gov.cn/zwgk/zcfb/art/2024/art_a4711acb06364199a3c5a06d7f2be6d8.html
4. Ministry of Commerce of China (2024). Announcement 2024 No. 46 on strengthening export controls on dual-use items to the United States (in Chinese). exportcontrol.mofcom.gov.cn/article/zcfg/gnzcfg/zcfggzqd/202412/1072.html
5. Ministry of Commerce of China (2025). Announcement 2025 No. 72, suspending clause 2 of Announcement 2024 No. 46 until 27 November 2026 (in Chinese). www.mofcom.gov.cn/zcfb/blgg/gg/2025/art/2025/art_bc4513421bb24faaa84e44c2e4f36dc5.html
6. Ministry of Commerce and General Administration of Customs of China (2025). Announcement 2025 No. 10 on export controls on tungsten-, tellurium-, bismuth-, molybdenum- and indium-related items (in Chinese). www.mofcom.gov.cn/zwgk/zcfb/art/2025/art_e623090907fc4e1092f0a4db72f57b95.html
7. Ministry of Commerce and General Administration of Customs of China (2025). Announcement 2025 No. 18 on export controls on certain medium and heavy rare-earth items (in Chinese). www.mofcom.gov.cn/zwgk/zcfb/art/2025/art_9c2108ccaf754f22a34abab2fedaa944.html
8. Ministry of Commerce and General Administration of Customs of China (2025). Announcement 2025 No. 70, suspending Announcements 2025 Nos. 55-58, 61 and 62 until 10 November 2026 (in Chinese). www.mofcom.gov.cn/zwgk/zcfb/art/2025/art_b1ec77dd3f0d4762952904df7cdaadec.html
9. Ministry of Commerce and General Administration of Customs of China (2025). Announcement 2025 No. 57 on export controls on certain further medium and heavy rare-earth items (in Chinese). www.mofcom.gov.cn/zcfb/blgg/gg/2025/art/2025/art_9510f263c3c246d8936fbfa684a58ca8.html
10. Ministry of Commerce and General Administration of Customs of China (2025). Announcement 2025 No. 58 on export controls on lithium batteries and artificial-graphite anode materials (in Chinese). www.mofcom.gov.cn/zcfb/blgg/gg/2025/art/2025/art_8ef8c6bf57e3437e826fcdf1c469aff8.html
11. Seoul Economic Daily (2026). "Korea Zinc's Onsan Smelter Produces 10 Tons of Antimony Daily for U.S. Defense," 8 March 2026. en.sedaily.com/finance/2026/03/08/korea-zincs-onsan-smelter-produces-10-tons-daily-of-defense
12. US Geological Survey (2025). "Antimony." Mineral Commodity Summaries 2025. pubs.usgs.gov/periodicals/mcs2025/mcs2025-antimony.pdf
13. US Geological Survey (2025). "Bismuth." Mineral Commodity Summaries 2025. pubs.usgs.gov/periodicals/mcs2025/mcs2025-bismuth.pdf
14. Newey, W. K. and West, K. D. (1987). "A simple, positive semi-definite, heteroskedasticity and autocorrelation consistent covariance matrix." Econometrica 55(3): 703–708. doi.org/10.2307/1913610
15. Bellemare, M. F. and Wichman, C. J. (2020). "Elasticities and the inverse hyperbolic sine transformation." Oxford Bulletin of Economics and Statistics 82(1): 50–61. doi.org/10.1111/obes.12325
16. Chen, J. and Roth, J. (2024). "Logs with zeros? Some problems and solutions." The Quarterly Journal of Economics 139(2): 891–936. doi.org/10.1093/qje/qjad054

Data

1. EU imports. Eurostat Comext, monthly detailed trade by CN8 code, imports of the EU27 from outside the EU, the United Kingdom excluded throughout; January 2021 to Jul 2026. Reused under the Commission’s reuse policy (CC BY 4.0). Bulk download.
2. Transit candidates' imports. UN Comtrade, monthly imports as reported by Malaysia and the Republic of Korea, partner China; derived figures only, not redistributed. comtradeplus.un.org.
3. World production. British Geological Survey, World Mineral Statistics, as held in the atlas, mean 2021–2023. bgs.ac.uk.
4. US imports. US Census Bureau international trade API, general imports by HS10 code (a US government work). census.gov.

Filing and code. The design was committed before any trade value was pulled, and every change after it is logged with its date: export-controls/PREREGISTRATION.md, and the follow-up on origins, Amendment A. Built by build_export_controls.py from out/export_controls.json, written by export-controls/analysis.py.