Criticality and geopolitics: Concentration and companionality: the anatomy of supply risk

3.1 The Herfindahl index turns market shares into one number

The Herfindahl-Hirschman index (HHI) is the sum of the squared market shares of all producers. With shares in percent it runs from near 0, for many tiny producers, to 10,000 for a monopoly. Squaring gives large producers disproportionate weight. A handy reading: 10,000 divided by the HHI is the number of equal-sized producers that would give the same concentration.

The tiles supply worked examples from World Mining Data's 2024 country shares of mine production. Gallium: China's 98.7%, squared, is 9,738; Russia, Japan and Korea add almost nothing, so the HHI is 9,739, equivalent to one producer. Rare earths: China's 67.8% gives 4,602, the United States (11.4%) adds 131, Myanmar (9.9%) 98 and Australia (5.1%) 26, for 4,862, about two equivalent producers. Cobalt: 5,706, about 1.75. Graphite: China's 75.9% contributes 5,765 of a 5,861 total. Compare copper at 1,089, about nine equivalent producers, crude oil at about 850 and wheat at 777 (FAOSTAT, 2024).

Thresholds come from competition policy. The 1982 US merger guidelines, which Tilton and Guzmán describe, treated markets below 1,000 as unconcentrated and above 1,800 as concentrated. Nassar and colleagues treat 2,500 as highly concentrated and found 43 of 62 metals above it.

Two cautions apply. A country HHI measures geography, not market power: Tilton and Guzmán stress that producers need the incentive as well as the ability to raise prices, since higher prices invite entry and substitution. And ownership can differ from geography. De Beers controlled about half of world diamond output through its marketing arm while its parent's share of production was 22%. The EU study raises the same ownership concern for nickel.

• World Mining Data 2026, country shares of mine production web ↗ · 2024, as shown on the gallium, rare earths, cobalt, graphite, copper and crude oil tiles
• FAOSTAT (2026), crop production by country web ↗ · wheat, 2024, as shown on the wheat tile
• Tilton & Guzmán (2016), Mineral Economics and Policy · pdf pp. 177-183, 191-192
• Nassar, Graedel & Harper (2015), By-product metals, Science Advances web ↗ · pdf pp. 3-5
• European Commission (2023), Study on the Critical Raw Materials for the EU 2023 web ↗ · pdf p. 40

3.2 The choke point is often processing, not the mine

Ore is only the first step. Separation, refining and chemical conversion are often more concentrated than mining, which is why the EU scores both stages.

Rare earths are the clearest case. On World Mining Data, China mined about 68% of world output in 2024, but the EU study puts its share of processing at 85% for the light rare earths and 100% for the heavy ones, including dysprosium and terbium. Lithium is mined mainly in Australia (38% in 2024), China (21%) and Chile (20%), yet China did 56% of processing in 2016-2020. Trade data show the pattern: in 2024 China took 69% of world lithium carbonate imports by value and supplied 80% of lithium hydroxide exports. Congo (Kinshasa) mined 63% of cobalt in the EU's reference period, while China, the leading refiner, took 61% of world imports of cobalt mattes, intermediates and metal (HS 8105) in 2024. China also took 63% of copper concentrate imports.

Processing positions are built, not found. Pitron describes how the French group Rhône-Poulenc, a leading rare earth refiner at La Rochelle, handed low-end refining and know-how to Chinese partners while believing itself decades ahead; by 2001 Chinese refiners had caught up. In the late 1990s Japan, the US and Europe made 90% of rare earth magnets; China later made three-quarters. Export quotas then gave foreign magnet makers a hard choice: stay at home and risk running short, or move to China.

Shares can move fast when a producer changes policy. Indonesia supplied 26% of nickel ore in the EU's 2016-2020 data. It banned ore exports in 2020 to build its own processing and battery industry, and by 2024 it mined 62% of world nickel. That concentration was created by industrial policy, not geology.

• European Commission (2023), Study on the Critical Raw Materials for the EU 2023 web ↗ · pdf pp. 10, 21, 27-28, 40
• World Mining Data 2026, country shares of mine production web ↗ · 2024, rare earths, lithium and nickel tiles
• UN Comtrade, export and import shares by value web ↗ · 2024, HS 283691, 282520, 8105 and 2603 on the lithium, cobalt and copper tiles
• Pitron (2020), The Rare Metals War · pdf pp. 87-89
• USGS Mineral Commodity Summaries 2026, Significant Events, Trends, and Issues web ↗ · pdf p. 20
• Scheyder (2024), The War Below · pdf p. 19

3.3 By-product supply cannot answer its own price

Many critical elements have no mines of their own. Gallium averages about 50 parts per million in bauxite and is recovered from the caustic liquor of alumina refineries, with a smaller share from zinc residues. Germanium comes from certain zinc and lead-zinc-copper ores and from lignite coal. Cobalt comes mainly from the copper deposits of Congo and Zambia and from nickel ores. Rare earths at Bayan Obo come out with iron ore, which Scheyder notes gave China a cost advantage.

Nassar, Graedel and Harper call the share of a metal obtained as a by-product its companionality. Of 62 metals, 38 are more than 50% companions. The tiles carry the estimates: gallium and germanium about 95%, rare earths 90%, cobalt 85%, copper 15%, nickel 5%.

Tilton and Guzmán explain the economics. By-product supply is capped by the amount contained in the host ore actually mined. Below the cap, supply can be fairly elastic, because only the costs specific to recovering the by-product count; the main product bears the joint costs of mining and milling. Near the cap the supply curve turns vertical, because a higher by-product price does not make anyone mine more host ore. The cap shifts with the host metal's market, so the by-product's supply depends on the price of something else.

There are escape routes. Recovery rates can rise: indium output rose 1,675% between 1975 and 2012 while zinc output rose 231%, because refiners recovered more indium from the same ore. Recovery can also be added at more plants. Both need the host producer to invest for a small revenue stream, and recovery rates for some companions are as low as 10%. These metals are also the least recycled: of the 31 with extremely concentrated production, 20 are recycled at under 1%.

• USGS Mineral Commodity Summaries 2026, Gallium web ↗ · pdf pp. 86-87
• USGS Mineral Commodity Summaries 2026, Germanium web ↗ · pdf p. 93
• USGS Mineral Commodity Summaries 2026, Cobalt web ↗ · pdf p. 75
• Scheyder (2024), The War Below · pdf p. 112
• Nassar, Graedel & Harper (2015), By-product metals, Science Advances web ↗ · pdf pp. 1-7
• Tilton & Guzmán (2016), Mineral Economics and Policy · pdf pp. 75-77

3.4 Across 32 studies, minor metals and by-products dominate the consensus

Lists differ in method and purpose, so Hayes and McCullough set methods aside and counted results. They compiled 32 studies published from 2005 to 2018, each assessing at least ten elements, and recorded how often each element was judged critical when assessed.

Three came out on top: the rare earth elements, the platinum group metals and indium. More than half the studies also named tungsten, germanium, cobalt, niobium, tantalum, gallium and antimony. The 11 studies published after 2014 added bismuth. Weighting regions equally brought in magnesium, abundant in the crust but produced mainly in China.

By-products stand out. The 17 elements usually recovered as by-products were judged critical in 68% of the studies that assessed them, against 44% for the 40 primary products, a statistically significant gap. The big host metals rarely count. Iron was judged critical in 20% of assessments, lead in 13%, zinc in 21%, nickel in 19% and copper in 14%, while cobalt, recovered from nickel and copper, scored 78%.

The consensus also reflects who writes the studies. Eighteen of the 32 came from the EU and the US, economies with large technology sectors. Australia, a mineral exporter with a small manufacturing base, counted only potash and phosphorus. Criticality moves with markets too: tellurium was named in 44% of all studies but 29% of recent ones, after silicon solar panels outgrew cadmium-telluride panels and tellurium became oversupplied. Lists also grow: the average study assessed 28 elements and found 15 critical, while studies since 2014 averaged 36 and 18. The authors expect the paradigm to widen towards fertiliser and industrial minerals as countries with different needs publish lists.

• Hayes & McCullough (2018), Critical minerals: a review, Resources Policy web ↗ · pdf pp. 1-7

See it in the table

Check your understanding

A material has two producing countries with 70% and 30% of supply. What is its HHI?

Why does a higher gallium price do little to raise gallium supply quickly?

In Hayes and McCullough's compilation, how often were copper and cobalt judged critical when assessed?