Is the ban leaking? India's non-basmati white rice prohibition (Jul-2023 → Sep-2024)

Trade-flow companion to the price wedge (R72). This is a DUAL-SCORE / alternative-track signal — never folded into any Tier-1 exposure score. The divergence between what a controlling state says it stopped selling and what actually kept leaving its ports IS the signal. Research, not investment advice; scope-shifting and origin-relabelling are INFERRED from customs composition plus the issuing government's own published findings, never asserted as smuggling on any single shipment.

Verdict

On 20 July 2023 India's DGFT (Notification 20/2023) moved non-basmati white rice from "Free" to "Prohibited" with immediate effect. India is the world's largest rice exporter; this is the single largest staple-food export control in the corpus. It leaked — and it leaked in three different ways at once, each one determined by where the drafter drew the scope line relative to a tariff line.

That makes this the corpus's cleanest test of the mode-F visibility rule (see circumvention-signals.md), because one instrument family produced all three boundary placements simultaneously, on one origin, in one window:

All figures: UN Comtrade, reporter India (699), annual, flow=export, HS 1006.20 / 1006.30 / 1006.40, canonical-row selection (max per reporter×partner×cmd×flow×period) to defeat Comtrade's dimension replication.

Read the controlled line alone (1006.30) and you would conclude the ban did nothing at all — exports rose to a record $11.1 bn while prohibited. Read the sibling lines and the mechanism is obvious.

The three leaks

1. Inside the line — and the government said so itself

The strongest evidence here is not ours. It is the Ministry of Commerce & Industry's own press release of 27 August 2023, five weeks after the ban, which states verbatim:

> "The Government has received credible field reports regarding misclassification > and illegal export of non-basmati white rice, export of which has been > prohibited with effect from 20th July 2023. It has been reported that > non-basmati white rice is being exported under the HS codes of parboiled rice > and Basmati rice."

The same release supplies the government's own numbers for the period to 17 August 2023, versus the year before:

The banned variety's own exports rose 4.36%. India's response was not to widen the prohibition but to attack the price tell: APEDA was instructed to register basmati export contracts only at USD 1,200/MT and above, holding anything below that in abeyance for committee review — because, in the Ministry's words, the lowest basmati contract price observed was USD 359/MT against an average basmati export price of USD 1,214/MT that month.

That is the regulator independently deriving, and acting on, the detector this case proposes below.

2. Outside the line — husked (brown) rice, the ×26 leak

Non-basmati white rice is milled. Husked (brown) rice — HS 1006.20 — is not "semi-milled or wholly milled" and therefore sat entirely outside the prohibition's scope description. A buyer can mill it on arrival.

India's brown-rice export line was dormant before the ban (16 kt in 2022, and ~0.1 kt/month through early 2023). It went to 123 kt in 2023 and 417 kt in 2024 — and then collapsed back to 41 kt in 2025 once the prohibition was lifted.

Where it went is the tell:

Viet Nam is the world's third-largest rice exporter. A top-three exporter buying 317 kt of another country's semi-processed rice — having bought none at all two years earlier, and none at all the year after — is not a food-security import. It is a milling-margin trade that exists only while the control exists.

Reuters reported the mechanism contemporaneously (27–28 February 2024): Viet Nam imported "at least 200,000 metric tonnes of husked brown rice from India between December and February", its first such purchases "in decades", in order to "process the grain and export the refined, white variety".

3. Transformation-washing — real, and honestly, small

Leg 2 is a mode C (transformation-wash) channel: Indian rice leaves under an out-of-scope HS line, is milled in Viet Nam, and re-enters world trade as Vietnamese white rice with lawful Vietnamese origin. The corpus's input-line ratio test (from the hardwood-plywood C⁻ case: controlled-origin input into the surger ÷ the surger's own outbound shipments) applies directly.

It returns a small number, and we report it as small:

• 2023, the one year both sides are in Comtrade: India→Viet Nam brown rice

101.6 kt ≈ ~91 kt milled-equivalent (~90% milling yield) against Viet Nam's 7,134 kt of HS 1006.30 exports = ~1.3%.

• 2024 cannot be computed on the same basis: **Viet Nam did not report HS 1006 to

Comtrade for 2024 or 2025 (zero rows returned for both import and export flows). This is a reporting gap, not** a zero flow, and it sits on the most important year.

Compare hardwood plywood, where the same ratio ran VN 105% and the US Commerce Department subsequently made an affirmative circumvention finding. At ~1–3%, India→Viet Nam rice is a real but marginal channel. It does not support any claim that Viet Nam's 2024 rice export boom was laundered Indian rice — most of that boom is genuine substitution into the gap India's own ban created. The wash is a rounding error on Viet Nam's trade and a large share of India's brown-rice line; those two facts are both true and must be stated together.

One further honesty note on the mirror: in 2023, India reported exporting 101.6 kt of husked rice to Viet Nam while Viet Nam reported importing 13.9 kt from India — a ~7× mirror gap, running in the direction that understates the channel from the importer's side.

The switching test — new to this corpus

Every other case in this register observes a control that is still in force. This one was imposed and then withdrawn (prohibition lifted by DGFT Notification 31/2024 on 28 September 2024, replaced by a USD 490/t minimum export price, which was itself removed in late October 2024). That gives something no static cross-section can: an on/off switch on the instrument, with the detector watching.

The out-of-scope line tracks the switch almost exactly — 16 kt → 123 → 417 → 41 kt. A leak channel that rises 26× while a control is on and falls ~90% within months of it coming off is being driven by the instrument, not by demand, weather, or price trend. This is the closest thing to a controlled experiment the corpus has, and it is the main reason this case is worth its length.

Monthly data dates the switch to within weeks of both events. India's husked-rice line, HS 1006.20, monthly exports (kt):

2025 runs 12.4, 8.0, 4.4, 4.9, 7.0, 1.9 kt (Jan–Jun), with Viet Nam at ~0 throughout.

The line is dormant until the month the prohibition lands, ramps for five months to a ~74 kt/month plateau, and Viet Nam exits within six weeks of the repeal — 36.1 kt in September 2024, 0.5 kt in November. Two independent corroborations fall out of this series: it reproduces Reuters' contemporaneous reporting of "around 70,000 metric tons... every month" (Dec-2023 = 70.7 kt, Jan-2024 = 70.7 kt to Viet Nam) and of Viet Nam being ~95% of India's husked-rice exports (70.7/74.2 = 95.3% in both months), neither of which we took from the article. Summed, Dec-2023 + Jan-2024 + Feb-2024 = 213 kt, matching the reported "at least 200,000 metric tonnes... between December and February".

(Dec-2023 and Jan-2024 are near-identical at one decimal — 74,185 t and 74,197 t respectively — but are distinct observations, not a duplicated row; the raw netWgt values were re-pulled and checked.)

One residual deserves naming, because it is a different instrument entirely: the world line stays elevated in Nov–Dec 2024 (33.1, 35.4 kt) after Viet Nam leaves. That flow is Nigeria (32.5 and 35.0 kt in those two months), which restricts milled rice imports to protect domestic millers and therefore buys brown rice to mill locally — a buyer-side control producing the same upstream-form substitution on the same tariff line, and one that keeps running after India's export ban is gone. Two controls stack on one HS line here and are separable only by partner.

The unit-value series switches with it, and is the second new detector here:

The intra-line rule this yields: when a control's scope boundary cuts inside a tariff line, volume is blind but unit value is not. Through the ban, 1006.30's volume was flat (17.9 → 16.3 → 16.5 Mt) while its unit value rose +27% ($526 → $669) — the cheap prohibited variety leaving the line and expensive unrestricted varieties (basmati, parboiled) filling it. When the ban lifted, volume jumped to a record 20.4 Mt and unit value fell back to $566. Flat volume plus rising unit value plus reverting on repeal is the composition-shift fingerprint.

Contrast 1006.40, where volume collapsed and price rose: that is the scarcity fingerprint of a line that genuinely was severed. The price–volume pair distinguishes a real bite from a relabelling, and neither is visible from volume alone.

Where the "ban" was never a ban

The broken-rice line (1006.40) carries a caveat that matters for anyone scoring this instrument. Its collapse was not uniform across buyers:

India's prohibition carried a government-to-government food-security carve-out, and the flow data shows exactly who it was for: the purely commercial buyer (China) went to literally zero, while Senegal never stopped and Indonesia's 2024 volume exceeded its 2022 volume. Consistent with the superhard-materials lesson already in this register — classify the instrument's legal scope before reading the flow — a discretionary carve-out makes a prohibition behave like an allocation regime, not a ban.

Note also that the post-repeal recovery is real but partial and uneven: partner count snapped back 15 → 59, but volume did not (808 kt in 2025, ~21% of 2022), and China returned at only ~4% of its pre-ban level. Breadth recovered; the anchor customer did not. Do not read the 2025 partner-count rebound as the line being restored.

Why it matters for the buyer

1. A control can leak without any origin ever being falsified. Two of the three leaks here involve no false flag at all — the goods were correctly declared, under a different HS code, from the same country. Screening on country-of-origin catches none of it. 2. Never read the controlled tariff line alone. The plywood case's lesson was "read the input line". This case generalises it: read the whole HS4 family — the controlled subheading, its siblings, and the semi-processed form one step upstream. India's 1006.30 line looked healthy precisely because the ban was working on part of it. 3. The blind spot is symmetric. A Tier-1 exposure score built on 1006.30 would have shown India's rice supply to the world as undisrupted through 2023-24. It was disrupted — for the specific buyers of the specific banned variety, many of them low-income West African and South Asian importers — while the aggregate line hit a record on price and mix. Aggregates hid a distributional shock. 4. Staple food behaves like a critical mineral. This is the first staple-food, seller-side prohibition in the register, and it produced the same modes (F scope-shift, C transformation-wash) as gallium, plywood and steel. The modes are properties of how instruments are drafted against tariff schedules, not properties of any commodity class.

Method & honesty rails

• What is claimed: that exports migrated into unrestricted tariff lines while

the prohibition was in force, and receded when it was lifted. That the Indian government itself found and acted on misclassification into parboiled and basmati codes.

• What is NOT claimed: that any identified shipment was smuggled or

fraudulently declared. Shipping brown rice, parboiled rice or basmati was entirely lawful throughout — that is the point of a scope shift. The misclassification finding is the Government of India's, made about the trade in aggregate, not ours about any exporter.

• No ownership tell was traced. Unlike the antimony (Youngsun) or graphite

(BTR) cases, we did not identify a common-ownership pipe between an Indian exporter and a Vietnamese miller. We looked and did not find one in public registries; this is recorded as an open gap, not as absence of one.

• The key year has a data hole. Viet Nam did not report HS 1006 to Comtrade

for 2024–25, so the 2024 input-line ratio is uncomputable on a like-for-like basis and the Vietnamese mirror cannot corroborate India's 317.2 kt. Labelled, not filled.

• Milling-yield assumption: ~90% brown→milled conversion, used only to state

the ratio order-of-magnitude. Nothing in the verdict turns on it.

• Data lags and revisions. Comtrade annual figures are revised; 2025 is the

most recent full year and the most likely to move.

• Sources. UN Comtrade (free, keyless preview API) for all trade figures;

Press Information Bureau / Ministry of Commerce & Industry release 1952629 (27-Aug-2023) for the misclassification finding, the government's own export tonnages and the USD 359 vs USD 1,214/MT basmati contract spread; DGFT Notifications 20/2023 and 31/2024 for the prohibition and its withdrawal (as filed in docs/iptm/actions/2023-07-20-india-non-basmati-white-rice-export-ban.md); Reuters (27-Feb-2024, read via The Star syndication) for the contemporaneous Viet Nam re-export account. No internal or subscription analyst material is used anywhere in this case.

Reproducing the trade tables

`` https://comtradeapi.un.org/public/v1/preview/C/A/HS ?reporterCode=699&period=<YYYY>&cmdCode=100620|100630|100640&flowCode=X ` Reduce each response with canonical-row selection — **max** per (reporterCode, partnerCode, cmdCode, flowCode, period) — before aggregating. Summing the raw response multiplies each corridor by its partner2×mode-of-transport×customs-procedure combination count and will corrupt the partner ranking, not merely the magnitudes (see scripts/py/comtrade_dims.py`).