The origin gap: the refiner is not the source

Key findings

• The exporter is not the miner. In 17 of 32 materials (2024 reconciled trade) the leading exporter differs from the leading miner. Import-origin statistics, taken at face value, misidentify the chokepoint in a majority of critical materials.
• Pure refiner/hub effects are large. In 4 of 32 materials a country that mines under 5% of world supply nonetheless exports over 25% of it. Beryllium is the extreme: Kazakhstan exports ~89% and mines none; the actual lead miner is the United States.
• The illusion is not only China’s. For many materials China is both the lead miner and the lead exporter — its chokehold there is genuine, not an artefact. The materials where exporter and miner diverge are instead fronted by industrial refiners and entrepôts: Finland for Congolese cobalt, Germany for Iranian strontium, the United States for Congolese tantalum.
• The gap can also hide concentration. Where ore is laundered through several different refiners before it ships (DR Congo cobalt, Indonesian nickel), the trade ledger looks more diversified than the mine base actually is — so the correction reveals hidden upstream risk, not only phantom refiner dependence.

1 · The measure

For each material, in the same year, we compare two public shares — a country’s share of reconciled world trade, and its share of world mine output:

A large positive gap means a country sells far more of the material than it digs out of the ground — it is processing or trans-shipping someone else’s ore. It is the single number behind the atlas’s thesis: the refiner is not the source. The gap is not an accusation of fraud; some of it is legitimate value-add (refined cobalt chemicals are a different product from mined cobalt). It measures where customs would mislead you about origin, material by material.

2 · Where the gap is largest (2024)

3 · The twist: the correction cuts both ways

The intuitive story is “China hides behind refineries.” The data only half-supports it. For many materials China is both the lead miner and the lead exporter, so its dominance is real rather than an accounting effect. The materials where exporter and miner diverge are fronted by industrial refiners and entrepôts — a German strontium “supply” is Iranian rock; a Finnish cobalt “supply” is Congolese ore.

So the corrective deflates apparent dependence on refiner countries — but it also does the opposite. A genuinely concentrated upstream (DR Congo cobalt at 74% of mine output, Indonesian nickel, Chinese rare earths) is more concentrated than the diversified-looking trade ledger suggests, because the ore is split across several refiners before it ships. The trade ledger flatters diversification in both directions; only subtracting the mine layer tells you which risk you actually face.

4 · What this is and isn’t

• It is a like-for-like comparison of two public measures — CEPII-BACI reconciled trade shares and USGS mine-production shares — per material, every year 2018–2024.
• It isn’t a claim that the refiner adds no value, nor that the traded form equals the mined form. Some gap is legitimate; the measure locates where customs would mislead you about origin, not fraud.
• Caveat. Gallium, germanium and hafnium share one HS6 code (811292) and cannot be separated in trade; those rows are flagged throughout the atlas.

Method

Bilateral trade is reconciled from UN Comtrade via CEPII BACI (validated against the official series: top-1 exporter matched 25/30, mean 3.5% share error). Trade shares are computed on reconciled exports per HS6 mapped to each material; mine-production shares are taken from the USGS Mineral Commodity Summaries. The origin gap is their difference in percentage points, computed per material per year 2018–2024, with 2024 as the headline. A companion note (short form) documents why raw trade needs reconciling at all — much of the apparent flow of many materials is re-export theatre, and roughly half of two-sided flows disagree with their own mirror. Full methods and all data are open. Read each row as an overlay of two different measures, not one observed pipeline.