What “companionality” means, and the sources
Companionality = the approximate share of world production that arises as a by-product or co-product of a host commodity (0 = always mined for itself; 100 = never mined for itself). A high value means supply is inelastic to the material’s own price: more is produced only when the host is, regardless of how “critical” it becomes. We cross companionality with the atlas’s trade concentration (HHI) into a double-jeopardy reading — supply-inelastic and geographically concentrated.
Sources & caveats: companionality figures compiled from USGS Mineral Commodity Summaries 2024 and Nassar, Graedel & Alonso (2015, Science Advances), and recomputed from mine-level open data (ICMM Global Mining Dataset 2025) as a validation column below. Values are round approximations of a genuinely fuzzy quantity (by-product share shifts with price and deposit) — treat them as tiers, not decimals. HHI is trade-based (Comtrade/BACI), so it measures export concentration, not mine concentration. → companionality.json.
The double-jeopardy map
Each bubble is a material. Right = more of it is an unavoidable by-product (supply can’t respond to its own price). Up = more geographically concentrated trade. Bubble size = trade value. The top-right is the danger zone: inelastic and concentrated.
Every material, ranked by double-jeopardy
Double-jeopardy index = companionality × trade concentration (HHI). High = you can neither diversify the supplier nor scale the supply.
| Material | supply type | by-product % | recovered from | HHI | DJI |
|---|
Are these numbers right? A mine-level cross-check
The by-product shares above are compiled from the literature (USGS, Nassar 2015) — a fair objection is that they’re curated, not computed. So we recomputed companionality independently, from open mine-level data: for every mine in the ICMM Global Mining Dataset (2025), is the metal its primary product, or does it ride along as a secondary one? The by-product share falls straight out of the counts — and for the metals the whole thesis rests on, the two agree.
| Hostage metal | literature | from mine data | mines |
|---|
Why this matters, and what it spawns next
Conventional supply-risk scores treat every material as if a price signal could summon more of it. For the by-product tier that is simply false: no gallium price will build a gallium mine. This reframes mitigation — for hostage metals the levers are recovery yield at the host, stockpiling, and substitution, not new mines. It also seeds the next layers: a risk re-weighting that penalises companionality, a recovery-yield / recycling lens (secondary supply is the only elastic source for these metals), and a host-shock model — what an aluminium or zinc downturn does to the criticals riding on it. This atlas grows by letting each finding pose the next question.