Critical Materials Atlas
Time · 2002–2024

Twenty-two years, drawn

A continuous measured trade series back to 2002 means the concentration story is no longer a snapshot. Pick a material and watch its exporters — and its concentration — evolve. Then see China’s export share climb across the marquee chains in one picture.

Each line is a country’s share of world exports over 2002–2024; the dashed line tracks how concentrated the trade is overall.
Reading the chart

Solid lines are each country’s share of world exports of the material (left axis, %); the dashed line is the export-concentration Herfindahl (right axis, 0–1 — higher = more concentrated). Measured CEPII BACI, 2002–2024 (HS02 vintage through 2016, HS17 from 2017); nowcast years excluded. Shares are over the captured bilateral flows, as elsewhere in the atlas.

Material: export shares + concentration over time

What drove the 2002–2024 concentration change

Each country's contribution to the change in export-HHI (share²2024 − share²2002, ×100). Red = pushed concentration up (gained share); green = pushed it down. This is the Family-5 decomposition: it names who concentrated a market.

The rise of China, in one picture

China’s share of world exports, marquee materials, 2002–2024.

The origin gap over time

Origin gap = the top exporter’s world export share minus that country’s own mine share. Mine share is the current USGS snapshot held fixed, so this isolates how the trade map drifted from today’s mining reality — a rising line means the refiner/hub illusion widened. Bold = average across all 32 materials; thin lines = the materials with the widest gap today. The dotted vertical marks the 2017 HS-vintage join, where the level steps for nomenclature/coverage reasons, not signal — so read the level cautiously (an HS17-only run is the clean confirmation).

Computed from the per-year reconciled flows → trends.json. A broad code (e.g. magnets = all metal permanent magnets) carries that breadth across the series. Origin gap uses current USGS mine shares as a fixed reference.

Which trends are statistically real

The export-concentration (HHI) trend per material, tested rather than eyeballed: Mann–Kendall significance (Benjamini–Hochberg FDR-corrected across the 32 materials), Theil–Sen slope, the Pettitt structural-break year, and the country that drove the 2002–2024 HHI change. Sorted by slope. The standard critical-minerals literature plots concentration but rarely tests it — with T=23, read p-values as screening evidence. Computed by build_trends.py (scipy + statsmodels).

MaterialSen slope /yrMK p (FDR)breakΔHHI driver