Corrections
When a published page turns out to be wrong, it is fixed and the fix is listed here: what it said, what it says now, why it was wrong, and where the right figure comes from. This is separate from what changed in the data, which lists new and revised source data, and from what's new, which lists changes to the site.
Every figure on the site is computed from a named source or cited to one, and drafts are fact-checked line by line before they are published. Errors still get through, and some are found afterward by readers or by later checks. To report one, use the contact page.
It said: Weekly statistics said the world's known rare earths reserves would last about 219 years at 2025 mining rates, that China produced 35% of the world's titanium (ilmenite) in 2025, and that Indonesia holds 48% of the world's known nickel reserves.
It says now: About 193 years of rare earths reserves, 34% for China's titanium (ilmenite) output, and 44% of nickel reserves for Indonesia.
What was wrong: Rare earths: USGS reissued the chapter on 26 May 2026 with Brazil's reserves at 11 million tonnes and the world total above 75 million tonnes, where the February data release had 21 million and above 85 million (see usgs_overrides.yaml); the statistic was frozen before that was applied. Nickel and titanium: the shares were computed on a different world total than the one USGS prints (nickel: above 140 million tonnes; ilmenite: 9.4 million tonnes, rounded).
It said: Three weekly statistics said Chile was a net exporter of 25 of 96 commodities and an importer of 12 (net balance $39 billion), China an exporter of 20 and importer of 71 (-$808.7 billion), and the Netherlands an exporter of 18 and importer of 55 (-$58.4 billion). Five blog posts said the site tracks 96 commodities, and gave per-country counts from that period: Saudi Arabia 24 commodities, Australia 53 ($276.3 billion), Indonesia 34 ($89.9 billion) and a top-three production count of 16 of 90.
It says now: Chile: exporter of 29 of 109 and importer of 28 ($33.9 billion). China: exporter of 27 and importer of 78 (-$807.4 billion). The Netherlands: exporter of 19 and importer of 67 (-$59.1 billion). The posts say the site tracks 117 commodities; Saudi Arabia exports 27 of them, Australia 61 ($277.2 billion), Indonesia 36 ($90.0 billion), and Indonesia is a top-three producer of 17 of 110 commodities with production figures.
What was wrong: The weekly statistics were frozen before the net trade report was rebuilt from the full bilateral table (see net-trade-recomputed-2026-10-12), and the posts' counts were written when the site tracked 96 commodities.
It said: Net trade figures in 10 posts were recomputed after a pipeline fix. Australia: net exporter in 44 commodities and net importer in 16, combined surplus $264.1 billion, iron ore $87.4 billion, natural gas $49.0 billion, coffee $460 million. Chile: net exporter in 25 and importer in 12, net exports $39.0 billion, largest net imports natural gas $2.39 billion, coal $694.9 million, corn $579.7 million. Indonesia: tin net export 'the same' as its $1.65 billion of exports. Saudi Arabia: net exporter in 8 and importer in 47, combined surplus $168.8 billion, dates $441.6 million, largest net imports gold $8.1 billion, steel $1.9 billion, rice $1.9 billion, barley $868.1 million. Gold: Japan +$19.7 billion, India -$58.5 billion, Turkey -$15.2 billion. DR Congo: cobalt 'its only tracked net-exporter position of any size'. Copper, coffee and DR Congo posts: 96 (or 'roughly 96') commodities in the net trade report; coffee described as at the extreme end with a handful of exporting countries.
It says now: Australia: net exporter in 51 and net importer in 20, combined surplus $265.1 billion, iron ore $87.3 billion, natural gas $48.8 billion, coffee $456 million. Chile: net exporter in 29 and importer in 28, net exports $44.4 billion, largest net imports crude oil $4.72 billion, natural gas $2.16 billion, coal $694.9 million (corn fourth at $493.4 million). Indonesia: tin net export $1.62 billion. Saudi Arabia: net exporter in 8 and importer in 59, combined surplus $170.1 billion, copper $635.7 million and dates $440.6 million, largest net imports gold $6.1 billion, rice $1.9 billion, steel $1.7 billion, sugar $1.5 billion (barley $868.1 million, tenth). Gold: Japan +$19.2 billion, India -$58.0 billion, Turkey -$10.8 billion. DR Congo: cobalt is its largest net export at $3.05 billion, with copper ore a near-equal $3.03 billion. The net trade report covers 109 commodities; coffee has 35 net exporting and 71 net importing countries.
What was wrong: The net trade report had been built from a table that kept only each flow's largest countries, so importer totals fell short and some countries' net positions were overstated or missing. It is now built from the full bilateral table, and the posts' figures were recomputed from it. A line counts as a net export or net import when the balance is at least $10 million. Figures that did not change at the precision shown (for example Chile's copper $30.9 billion, Indonesia's coal $30.9 billion and palm oil $21.1 billion, Vietnam's rice net of $2.4 billion, the five largest net crude exporters and importers, the five largest net rice importers and gold's five largest net exporters and importers by country) were left as they were. The DR Congo sentence also changed because the earlier claim that cobalt was its only net export of size did not hold against copper ore.
It said: The phosphate fertilizers page's 'who sells to whom' chart covered only superphosphates (HS 3103, $3.5 billion in 2024) and left out ammoniated and compound fertilizers (HS 3105, $37.6 billion); an HS line was labeled diammonium phosphate (DAP) when HS 310540 is monoammonium phosphate and its mixtures with DAP. The phosphate rock and logs pages showed no China or Russia in their trade charts.
It says now: The chart counts HS 3103 and 3105 together ($41.2 billion). The phosphate rock page shows China's place in the fertilizer trade (China mines 44.0% of phosphate rock but is 15th in rock exports and second in fertilizer exports), and the logs page shows Russia's place in sawnwood (third, 9.4%, against 18th in log exports).
What was wrong: The chart used only the page's primary customs line, and the leading producers sell processed products on other lines.
It said: Chile sold $30.9 billion of copper ore and concentrate, 29.2% of world export value; Peru 19.3%; South Korea bought $4.1 billion.
It says now: Chile $31.0 billion, 29.3%; Peru 19.4%; South Korea $4.7 billion.
What was wrong: Three values did not match the trade table the post cites. The Korea figure was off by about 15%.
It said: China exports little primary aluminum because of a 15% export tax; smelting takes roughly 14 megawatt-hours a tonne; about four to five tonnes of bauxite make two tonnes of alumina; the US Section 232 tariff rose to 25% in February 2025.
It says now: China levies a 30% export duty on unalloyed unwrought aluminum (HS 7601.10) and a 15% provisional duty on alloys; smelting took about 13 to 14 megawatt-hours a tonne worldwide in 2022; USGS gives about four tonnes of dried bauxite for two of alumina for one of aluminum; the 25% tariff began in March 2025.
What was wrong: The 15% rate applies to aluminum alloys, not to unalloyed metal; the electricity and bauxite figures had no source and the bauxite ratio disagreed with USGS; the tariff step was misdated.
It said: The oranges page's trade tables and its 'who sells oranges to whom' chart did not include Brazil, the largest grower, and the orange juice page's flows counted only not-frozen juice.
It says now: The oranges page explains the gap and shows the orange juice trade beside the fresh-fruit trade; the juice flows now count frozen concentrate (world $4.6 billion in 2024, Brazil 45%) as well as not-frozen juice.
What was wrong: Brazil squeezes most of its oranges and exports juice, which is a different customs line, and the juice chart left out frozen concentrate, Brazil's largest juice line (HS 200911, $2.8 billion of world exports in 2024).
It said: The rice price chart drew a broken line where the 25% broken rice series has no value (1988 to 1990).
It says now: Months with no value are skipped.
What was wrong: A missing month was passed to the chart as a number, which wrote an invalid point into the line.