Market transmission case study — China's 2024-12-03 Ga/Ge/Sb export ban to the US
Template case for the per-event qualitative layer (2026-07-05). The mechanical directional event study scored this event WRONG-SIGNED (producer−consumer spread −9.9% t+20, −49% t+60; docs/calibration/EVENT_STUDY_DIRECTIONAL.md). A name-by-name reconstruction shows the hypothesis was RIGHT — the alternatives to Chinese supply re-rated violently — but the cohort machinery could not see it. Research, not investment advice. Price moves verified against Yahoo daily closes (baseline = 2024-12-02 close); channels verified against contemporaneous reporting.
Verdict
The tradable expression of this event was 3–5 small-cap ex-China alternatives plus one structural short, not a cohort average:
| # | expression | move | channel |
|---|---|---|---|
| 1 | UAMY long (United States Antimony) | +38% day 1, +168% in 4 sessions, +210% by Mar-31 | only US antimony smelter; re-prices 1:1 with the Rotterdam Sb quote the ban squeezed |
| 2 | PPTA long (Perpetua Resources) | +17% day 1, +36% by Dec 6; −22% on the Feb-2025 $71M equity raise (the exit signal), +18% net Q1 | only US antimony reserve (Stibnite), DoD-backed; Dec-9 Sunshine Silver processing MOU |
| 3 | AXTI short (AXT Inc) | −11% day 1, −36% by Q1-end (−31% single day on Feb-20 results) | US-listed but its GaAs/Ge substrates are made INSIDE China (Tongmei) — now behind MOFCOM permits; China added InP on 2025-02-04, gating all three product lines |
| 4 | LRV.AX long, 3-month horizon (Larvotto) | fell through Dec, then ~+100% Jan–Mar | Hillgrove = Australia's largest antimony development; re-rated with the spot squeeze, not the headline |
| 5 | AII.TO long as the sequel (Almonty) | flat Dec 3; +130% when China ran the identical playbook on tungsten (2025-02-04) | the ban made the follow-on event class predictable |
Spot context: antimony ~$38–39k/t pre-ban (from ~$13k early 2024) → $50k+/t H1-2025; germanium record $2,900–3,000/kg (Rotterdam, Dec 6); gallium initially FLAT (direct CN→US flows already zero since Aug-2023 licensing — the ban was symbolic on flows) then ~+141% over 2025.
Why the mechanical screen got it backwards — five failure modes
1. Register data gap (the biggest): the action file listed only germanium-gallium in target_materials — antimony was missing despite being in the action title. UAMY and PPTA, the two largest winners of the episode, were structurally invisible to the backtest. (Fixed 2026-07-05: antimony added to the action file.) 2. Asset location beats domicile: AXTI matches "US company + gallium exposure" and looks like a beneficiary; its production is in Beijing, so a US-targeted ban makes it the highest-conviction loser. Domicile-based producer classification inverts the sign. 3. Micro-cap dilution: the true winners are $100–500M names our dossier universe under-covers; equal-weighting them against diversified mega-caps (Teck flat — Trail germanium immaterial to consolidated earnings) averages the signal to noise. 4. Confounders in tiny cohorts: Korea Zinc +42% that week was the MBK control battle + Korean martial-law shock, not germanium. With n_prod=4, one confounded name dominates the mean; the −49% t+60 spread is cohort noise, not information. 5. Consumers don't lose: Ge/Ga are trace inputs (<1% COGS) for Coherent/Teledyne/Qorvo — they stockpile and pass through. QRVO −0.5%, SWKS −1.0% on Dec 3. Input criticality ≠ P&L materiality; this is why the consumer-short leg fails in every statistical test.
Reusable transmission map (apply to each new event)
- Ch.1 — scarcity premium, ex-China pure-play (the reliable winner class):
find the 1–3 listed ALTERNATIVES to the cut supply — smallest, purest first. Screen: ex-China producer/processor/developer of the exact material, listed, <$1B mcap preferred. This is where the +100–200% lives.
- Ch.2 — domicile trap (the short): listed companies whose exposure to the
banned material sits in assets INSIDE the restricting country.
- Ch.3 — already-priced check: was there a prior escalation (licensing
before ban)? If flows were already zero, the headline is symbolic — trade the spot-price expression (Ch.1 over 3 months), not the announcement.
- Ch.4 — skip the consumer short unless the material is >5% of COGS with
no inventory buffer (rare).
- Ch.5 — the sequel: each event makes the next event class predictable
(Ga/Ge/Sb Dec-2024 → tungsten Feb-2025 → Almonty +130%). Watchlist the alternatives for the next material China controls (our register + upcoming queue is exactly this list).
- Exit signature: the pure-play monetizes the pop with an equity raise
(PPTA, Feb-2025 −22%) — that is the cycle-end marker.
Product implication
The sellable per-event artifact is THIS document, produced within days of each new event: named alternatives with channels, the domicile trap, the sequel watchlist — with the statistical layer as honest context (n too small for a mechanical basket; case-study depth is the edge). Sources: MINING.COM, CSIS, Fastmarkets (Ge/Ga spot), AXT 10-K/8-K (SEC), StockTitan (PPTA MOU), strategicmetalsinvest.com price histories; all price moves cross-checked on Yahoo Finance daily closes.