Barite: the flag that reports nothing

Trade-flow companion to the price wedge (R72). DUAL-SCORE / alternative-track signal — never folded into any Tier-1 exposure score. Origin-relabelling is INFERRED from statistical implausibility, never asserted as smuggling on any single shipment. Research, not investment advice.

Verdict

Three findings, and only one of them is the one the anchor control predicted.

1. The anchored control has not bitten yet — say so, don't dress it up. India's DGFT reclassified Grade A/B baryte from Free to Restricted on 17 April 2026. Recomputed 2026-10-02 on a matched 7-month window (Jan–Jul, one month more than the original H1 read): US imports from India are up 17.9% year-on-year (404,570 t → 476,877 t) at flat unit values — softer than the +50% the H1-only figure showed, because July-2025 alone carried 96,140 t to the US, the single-month lumpiness the original caveat below warned about. Direction unchanged: no collapse, no price signature of a grade downgrade. Either DGFT authorisations are being granted routinely (it is a licence regime, not a ban) or the window is still too short against vessel lumpiness. Register-state, not world-state: this is "no effect detected yet", not "no effect".

2. The real anomaly in this material predates the control and is large. Roughly 8–15% of all US barite imports (8.4% in 2023, 14.6% in 2025, 13.9% in H1-2026) arrive under a Vietnamese flag that neither Vietnam's mine production nor Vietnam's own customs return can account for. In 2023 — the last year Vietnam filed HS 251110 to Comtrade — Vietnam reported exporting 19,897 t of barite to the entire world and 0 t to the United States, while the United States reported receiving 204,116 t from Viet Nam. A 10.3× mirror gap against a line the reporter says is zero.

3. China's leg is substitution, not laundering. Chinese direct shipments to the US fell −66% H1-on-H1 with rising unit values, and the gap was filled by Morocco — a genuine 1,000 kt/yr producer — not by a phantom flag. Recomputed 2026-10-02, matched Jan–Jul window: China −74.3% (145,607 t → 37,424 t), Morocco +117.0% (130,262 t → 282,705 t) — same direction, different magnitude; Morocco's July-2025 base was itself unusually low (64,522 t that single month), so the matched-window move is smaller than the H1-only +300% but the substitution story holds. Same shape as the bismuth case: the squeeze bit, and an ally with real capacity took the volume.

Upstream origin of the Vietnamese line is NOT traced. That is why this case is GATE 0, not GATE 1. What is missing is named at the bottom.

Why barite is worth watching at all

Barite (BaSO₄) is the weighting agent in oil- and gas-drilling fluid; USGS states flatly that "there are no large-scale alternatives to barite in oil- and gas-drilling fluids." The US withholds its own production as proprietary and is overwhelmingly import-dependent. The API drilling specification turns on specific gravity — the 4.2 SG standard, with an alternate 4.1 SG weighting-agent specification issued in 2010 as global 4.2 reserves dwindled. That SG threshold is exactly the line India's notification cuts along, which is what makes this control interesting: it is a grade-line control, not a country-line control.

Layer 1 — volume implausibility: Viet Nam

USGS Mineral Commodity Summaries names eleven producing countries and lumps the rest into a single residual. Viet Nam is not separately listed in either the 2025 or the 2026 edition.

Source: USGS Mineral Commodity Summaries 2026 (Feb-2026), Barite chapter, p. 49 — https://pubs.usgs.gov/periodicals/mcs2026/mcs2026-barite.pdf. 2025 figures estimated. The 2025 edition (Jan-2025) gives the same eleven-country list with "Other countries" at 314/320 kt for 2023/2024.

Against that residual, the US-reported Vietnamese line:

Source: UN Comtrade, reporter USA (842), flow M, HS 251110, monthly, net weight in tonnes; free public preview + subscription API. Single dimension combination (partner2=0, mot=0, customs=C00) — no replication de-duplication needed on this reporter, unlike the Indian and Vietnamese returns below. Recomputed 2026-10-02: latest month available is now 2026-07 (was 2026-06), added as the "Jan-Jul 2026" column — "H1-2026" below now contains ~2.5 months of post-control data, "Jan-Jul 2026" ~3.5 months. Vietnam's share of the Jan-Jul 2026 total is 13.5 %, essentially unchanged from the H1-2026 figure of 13.9 % — the instability recomputing exposed is concentrated in the YoY move figures (Verdict 1 and 3), not in Vietnam's share.

The test. In 2025 the United States alone recorded 307,074 t of barite of Vietnamese origin — 88% of the entire world "Other countries" residual (350 kt). For that to be genuine Vietnamese mine output, Viet Nam would have to be essentially the whole residual, ship almost all of it to one customer, consume none domestically, and leave ~43 kt for every other unlisted producer on earth combined (Thailand, Bulgaria, Peru, Algeria, Myanmar, Nigeria, Germany…). Barite is a low-value bulk mineral at ~$110–130/t; nobody freights it around the world for fun.

Counter-hypothesis, stated because it is live: industry trade press treats Viet Nam as a genuine barite origin, and Vietnamese deposits (Tuyên Quang, Lạng Sơn, Cao Bằng, Thanh Hóa) are real. If Vietnamese output is simply uncounted by USGS, the residual is wrong rather than the flag. This case does not claim to have settled that — see "what would settle it".

Layer 2 — the customs mirror: a reporter that reports zero

Viet Nam last filed HS 251110 to UN Comtrade for 2023. Its own return:

Source: UN Comtrade, reporter Viet Nam (704), HS 251110, 2023, both flows, free public preview API. De-duplicated: the Vietnamese and Indian returns carry multiple aggregation rows per (period, partner) across the customs/mode-of-transport dimensions; summing them double-counts. The top-level aggregate per (period, partner) is taken — max, never sum.

Two readings, and the honest answer is that both are partly true:

• Vietnam's customs return is incomplete. China reports exporting 10,299 t of

barite to Viet Nam in 2023; Viet Nam records receiving 7,321 t. The land border is under-captured by ~29% on a line both sides agree exists. A return that misses a third of a recorded border flow may miss much more.

• But incompleteness cuts the other way too. Whatever is arriving in the US under

a Vietnamese flag, Viet Nam's statistical system does not see it leaving — not as export, not as re-export. A transit movement through a Vietnamese port that never legally enters Vietnamese commerce would produce exactly this signature, and would also mean the US "country of origin" field is recording a port, not an origin.

Layer 3 — the input-line test fails, and that is informative

The mode-C diagnostic from the hardwood-plywood and Xinjiang-cotton cases is: read the input line into the surger, not the finished-good origin line. Applied here:

Source: UN Comtrade, reporter China (156), flow X, HS 251110 (China's return carries one row per partner — no de-duplication required); US line as above.

Compare hardwood plywood, where the same ratio ran 105% for Viet Nam and Commerce convicted on it. Here it runs 5–9%. Reported Chinese feedstock cannot source the Vietnamese line — so the reflexive "it's laundered Chinese barite" story is not supported by the numbers. Either the feedstock is Laotian (landlocked, 250–260 kt/yr, exits through Vietnamese ports, and does not report to Comtrade at all), or it is genuinely Vietnamese and uncounted, or it crosses the Chinese land border unrecorded. Not traced. Labelled as untraced.

Layer 4 — the anchored control: India, 17-Apr-2026

DGFT Notification 12/2026-27 (Gazette S.O. 2222(E), 17-Apr-2026) moved Grade A (SG ≥ 4.2, ITC-HS 25111010) and Grade B (SG 4.10–4.20, 25111020) from Free to Restricted. Grade CDW (SG < 4.0, 25111090) stays free. India is the world's largest producer (3,000 kt in 2025) and the largest supplier to the US.

\July-2026 is excluded from the Jan-Jul unit-value range and flagged separately below — it is a genuine outlier, not folded into "flat".*

Recomputed 2026-10-02: the H1-only +50% overstated the move — July-2025 alone carried 96,140 t to the US from India, a single-month spike the original caveat below flagged as a live risk. The matched 7-month figure (+17.9%) is the current central number; India's share of total US imports is unchanged either way.

New in the recompute — a first, unconfirmed price signal. July-2026's India→US unit value is ≈$392/t, more than 3× every prior month in the series (range 78–132 through June-2026), on the lowest monthly volume of the whole window (15,122 t vs a typical 70–90k t/month). That is the shape a licensing-premium or a grade-mix shift (more Grade A/B at a higher declared value, less low-grade CDW) would leave — but it is one month, not a trend: could equally be a part-month data artifact or a handful of high-value contracted cargoes. Not enough to revise the Verdict's "no price signature" line; worth a dedicated re-check once 2026-Q3 data is complete.

The expected circumvention fingerprint for a grade-line control is not a transit country at all — it is an intra-line grade shift: Grade A/B tonnage re-declared into the free CDW line, same molecules, new sub-heading. That would be invisible at HS6 (all three grades sit inside 251110) and would normally leave a falling unit value. Neither signature is present: volume is up and unit value is flat. On the evidence available, the control is being administered as a licence regime that grants, not as a stoppage.

One loose thread, recorded not resolved: India's own reported exports to the US (de-duplicated: 1,286,583 t in 2023; 781,285 in 2024; 1,175,050 in 2025) run 1.21× / 1.19× / 1.62× the US-reported arrivals. The 2023–24 gaps are ordinary mirror noise (FOB/CIF, timing, free-trade-zone processing — USGS notes US general imports and imports-for-consumption differ for exactly this reason). The 2025 widening to 1.62× is not explained here.

Transmission chain

``` India (3,000 kt/yr, 42% of US supply) └─ DGFT Notification 12/2026-27, 17-Apr-2026: Grade A/B → Restricted, CDW free └─ [expected] intra-line grade shift into 25111090 → NOT DETECTED (needs 8-digit data) └─ [observed] US arrivals +50%, unit value flat → licences granted

China (2,200 kt/yr) └─ US tariff wall → direct US line −66% H1-on-H1, unit value UP └─ absorbed by Morocco (+300% H1-on-H1, genuine 1,000 kt producer) → SUBSTITUTION

"Viet Nam" (not a listed producer; own return: 19,897 t world exports, 0 t to US) └─ 204–307 kt/yr arriving in the US under a Vietnamese flag └─ feedstock: Chinese input line only 5–9% of it → UNTRACED └─ 88% of the entire world "other producers" residual → IMPLAUSIBLE AS ORIGIN ```

What this implies for the Tier-1 blind spot

A Tier-1 exposure score built on declared country-of-origin will read a US drilling-fluid buyer as diversified across India / Viet Nam / Morocco / Mexico. On these numbers the Vietnamese leg — 8–15% of US supply, 14.6% in 2025 — has no traceable origin at all. That is not a low-risk leg; it is an unmeasured leg, and per the standing rule, absent data is not good news. The correct treatment is to mark the Vietnamese share as origin-unverified in the alternative track and leave the Tier-1 score alone.

Caveats

• Inference, not proof. Nothing here identifies an entity, a shipment, or an act of

misdeclaration. A mirror gap and a residual-share ratio are statistical objects.

• Any single shipment is innocent. Vietnamese-origin barite certainly exists.
• The 2023 Vietnamese return is three years old and is the only one on file; 2024–25

Vietnamese imports could have risen enough to feed a genuine grinding industry, and nothing above rules that out.

• Grinding is not origin. Whether the Vietnamese line is crude or ground barite would

discriminate hard between relabelling and processing — and is not resolvable at HS6.

• 2.5–3.5 months of post-control data on a lumpy bulk trade (India's US line was

0 t in Apr-2026, 228,815 t in May-2026, 0 t in Jun-2026, 15,122 t in Jul-2026) is too thin for a verdict on the India control. Confirmed by the 2026-10-02 recompute: adding one month (Jul-2026) moved the YoY figure from +50% to +17.9% and Morocco's from +300% to +117%, both still correct in direction, both materially different in size. Re-run again after 2026-Q4 data lands.

• No price-line claim. Unit values here are customs value ÷ net weight, not

assessments.

What would settle it — the next three pulls, in priority order

1. India DGCI&S 8-digit exports for 25111010 / 25111020 / 25111090, monthly across Apr-2026. This is the only dataset that can see the grade shift the control invites. Comtrade stops at HS6 and cannot. 2. US Census HS-10 split (2511.10.10 crude vs 2511.10.50 ground) by country, via the free api.census.gov intltrade series — requires a free API key this VPS does not have (the unauthenticated call returns "Missing Key"). Crude-vs-ground on the Vietnamese line separates relabelling from a real grinding industry. 3. A named Vietnamese exporter of record, from Vietnam's National Business Registration Portal, and its ownership. No such entity was traced for this case; the common-ownership layer is empty and labelled empty.

Sources

• DGFT Notification No. 12/2026-27, 17-Apr-2026 — https://content.dgft.gov.in/Website/dgftprod/2dfe80e3-60f8-4986-9892-567a4438828e/Notification%2012%20(1).pdf
• USGS, Mineral Commodity Summaries 2026, Barite — https://pubs.usgs.gov/periodicals/mcs2026/mcs2026-barite.pdf
• USGS, Mineral Commodity Summaries 2025, Barite — https://pubs.usgs.gov/periodicals/mcs2025/mcs2025-barite.pdf
• UN Comtrade, HS 251110 — reporters USA (842) monthly 2023-01→2026-07 (recomputed

2026-10-02, extended from 2026-06); Viet Nam (704) 2023; China (156) 2023–24; India (699) 2023–25. https://comtradeapi.un.org/

Replication (free public preview endpoint, no key):

```bash curl -s "https://comtradeapi.un.org/public/v1/preview/C/A/HS?reporterCode=704&period=2023&cmdCode=251110&flowCode=X" curl -s "https://comtradeapi.un.org/public/v1/preview/C/A/HS?reporterCode=156&period=2024&cmdCode=251110&flowCode=X"

De-duplicate: group by (period, partnerCode), keep the partner2Code=0 / motCode=0 /

customsCode=C00 row. NEVER sum across those dimensions.

```

Replication for the 2026-10-02 recompute (subscription endpoint, needs COMTRADE_API_KEY; reuses scripts/py/comtrade_dims.dedupe_dimensions):

```bash curl -s "https://comtradeapi.un.org/data/v1/get/C/M/HS?reporterCode=842&period=202507&cmdCode=251110&flowCode=M&partnerCode=699,504,704,484,418,156" \ -H "Ocp-Apim-Subscription-Key: $COMTRADE_API_KEY"

Repeat per period 202501..202507 and 202601..202607; dedupe before summing.

```