Economy
From ore to GDP, sources, method, validation
Layer 0 measured the planet's physical base: each of the major resources the world extracts every year, priced at the point of extraction. Layer 1 asks the next question: once this material has been refined, manufactured, transported, and sold, how much of the world's economic output does it end up carrying?
The shortcut would be to take a resource's market value and scale it up by its sector's average multiplier. That shortcut was tried first, and rejected. A ton of sand and gravel sells for $15 and turns into concrete after almost no processing. Crude oil sells at a similar price per ton but turns into gasoline, plastic, fertilizer, and pharmaceutical intermediates - a much longer value chain. A flat sector-average multiplier erases exactly the difference the question is about.
The right answer means tracing each material through the real structure of the world economy: which sectors buy it, where those sectors sell it on to, and how far that chain extends before the value turns into someone's wage, profit, or tax.
1What it is built from
Every figure below ties back to a published, citable data set. Nothing in Layer 1 is a rough guess.
| Source | What it's used for | Year |
|---|---|---|
| OECD Inter-Country Input-Output (ICIO), 2025 edition | Full transaction matrix for 85 countries × 50 sectors: what each sector in each country bought from other sectors, plus total output and value added | 2022 |
| FAOSTAT, Value of Production domain | World production value by product (corn, wheat, cattle...) - an actual figure, not a derived estimate | 2024 |
| USGS Mineral Commodity Summaries | Reference unit prices for metal ores (iron, copper, gold) | 2025 |
| EIA international energy statistics | Reference unit prices for crude oil, natural gas, and coal | 2025 |
| IMF World Economic Outlook | World GDP figure used to carry the 2022 base forward to the current year | Apr 2026 |
2Method
Each of the world's 50 sectors falls into one of three roles. Layer 1's job is to dissolve the middle role into the other two - without losing or double-counting a single dollar.
7 · Resource sectors
Where a material first enters the economy: agriculture & livestock, forestry, fishing, coal, oil & gas, metal ores, quarrying.
30 · Transitional sectors
Manufacturing, construction, electricity/water, wholesale-retail trade, transport, food service. They add real value, but every input they touch can be traced back to a resource or forward to an end point.
13 · Pure information/services
Finance, real estate, health, education, public administration, telecom, and the like - sectors whose output doesn't carry forward as a physical good.
This is exactly where a naive approach breaks down: the 30 transitional sectors. Wholesale trade, for instance, sells the output of nearly every material on earth. Attribute its margin in full to every material it touches, and you count the same dollar seven times over. Ignore it, and you understate every material's true reach. The fix is to trace where that margin actually came from and return it - exactly, according to real purchasing patterns - to the 20 end points.
Map the 29 Layer 0 resources to their resource sectors
Corn, wheat, cattle, and nine other agriculture/livestock items are all grouped under a single OECD sector code (A01) - the classification doesn't split agriculture by product. Oil and gas share B06; iron, copper, gold, and uranium share B07.
Build the real purchasing map
From the raw ICIO transaction table, how much of each sector's cost comes from which other sector is extracted - not an assumption, but 2022's real trade flows aggregated from 85 countries into 50 world sectors.
Dissolve the 30 transitional sectors
Each transitional sector's margin is pushed backward through its own real purchasing pattern - however many steps that takes - repeatedly, until every cent lands in one of the 20 end points. This is an absorbing Markov chain: the 30 transitional sectors are the transient states, the 20 end points are where the walk is forced to stop (N=(I-Q)⁻¹, B=R·N). The result is exact - nothing is left hanging, nothing is double-counted.
Split each resource sector's new total back across 29 rows
After a resource sector (e.g. A01) receives its share from the transitional economy, that grown total is split back across the individual Layer 0 items that feed it (corn, wheat, cattle...) by their own share of production value - using FAOSTAT/USGS/EIA's own data wherever possible.
Add the tax bridge
Value added at basic prices isn't quite GDP - product taxes minus subsidies (VAT, excise, etc.) sit between the two. Since this can't be attributed to any single sector, it's added as one standalone line (+7,0 T USD).
Scale the 2022 structure to the current year
OECD hasn't yet published ICIO data past 2022. Every ratio in Layer 1 reflects 2022's real structure; only the overall scale is carried forward, using a single multiplier tied to the most recent known world GDP (×1,17 → 2025 total).
3What came out
| Block | Own VA | After dissolution | Multiplier |
|---|---|---|---|
| 7 resource sectors → 29 Layer 0 items | 7,98 | 24,16 | 3,0× |
| 13 pure service sectors | 44,93 | 69,64 | 1,6× |
| Tax bridge (TLS) | - | 7,01 | - |
| World GDP, 2022 structure | 93,81 (+TLS) | 100,81 | - |
Official 2022 world GDP (World Bank/FRED): 102,86 T USD - the 2,0% gap corresponds to the normal revision difference between data sources (see §5). Once scaled to the IMF's 2025 estimate (118,18 T USD), the same 43-row table lands exactly on the current-year total.
| Sector | Multiplier |
|---|---|
| Metal ores (iron, copper, gold, uranium, other) | 5,1× |
| Coal | 4,8× |
| Crude oil and natural gas | 4,3× |
| Other mining and quarrying (sand, limestone, clay, gypsum) | 3,6× |
| Forestry | 3,0× |
| Agriculture and livestock | 1,9× |
| Fishing | 1,7× |
Oil and metal ores carry the longest downstream chains - every barrel and every ton of ore passes through refining, manufacturing, and years of intermediate use before its value fully surfaces. Agriculture and livestock products, sold closer to their raw form, carry a shorter chain.
4How it was tested
Coal's implied price against the real market
ICIO's own coal sector output implies a mine-mouth price of $93/ton. China's 2025 mine-mouth thermal coal price is $80-90/ton - and China alone accounts for more than half of world coal production, so that's exactly the figure it should match, and it does.
Two independent routes to GDP arrive at the same place
Value added + tax bridge, built entirely from ICIO, comes within 2,0% of the World Bank's independently compiled 2022 GDP figure - within 0,3% for 2021. Two unrelated statistical pipelines converging this closely is strong evidence that neither has a structural error.
Nothing is lost or double-counted in the dissolution
By construction, the Markov absorption matrix's column sums are exactly 1,000000. Total value added before and after redistributing the 30 transitional sectors matches to six decimal places - not an approximation, a hard conservation check.
Agriculture weights checked against their own production value
The weight used to split agriculture's total across corn, wheat, rice, and so on is FAOSTAT's own Gross Production Value for each product - not a derived price - cross-checked row by row against FAOSTAT's own published world agriculture total.
5Where the limits are
Sector classification doesn't distinguish every material
ICIO's 50 sectors can't tell corn from rice, or iron from copper, at the transaction-matrix level - that distinction only exists in FAOSTAT/USGS's own production-value data, which is used wherever available. Items with no independent data (uranium; sand, gravel, and other quarry products; industrial roundwood and fuelwood) are split by their own AtSource value - a weaker but still principled alternative.
The base year is up to three years behind the current date
OECD ICIO's most recently published year is 2022. Everything about the table's structure - how much of the economy a material chain touches, relative to every other material - reflects that year. Only the overall scale is carried forward, using a single proportional multiplier tied to the most recent known world GDP - not a fresh reconstruction.
Forestry and construction materials still lack an independent production-value source
FAOSTAT's Value of Production domain doesn't cover forest products, and there's no equivalent source yet for the construction-minerals total. Both groups are split using Layer 0's own AtSource estimates until a better source turns up.
The absorption assumes 2022's purchasing pattern is the right routing
The Markov dissolution routes a transitional sector's margin backward through its real, observed 2022 purchasing mix. That's a defensible assumption, not a law of economics - a different year's mix, or a genuinely behavioral model of how firms would respond to a shock, could route the same dollar differently.
6Historical series (Historical Development page)
Layer 1's main work is a snapshot of the 2022 base year. A separate Historical Development page (/layer1/tarihsel) tracks population and total world GDP across 14 time points from 10,000 BCE to today. GDP is expressed as "2025 USD equivalent": today's nominal 118 T$ (Layer 1's own ICIO total) is the anchor, and past years are scaled back by the Maddison Project Database 2023 real-growth index. The Maddison series reaches 1 CE; earlier points rest on a bottom-up per-capita estimate and are effectively zero at this scale.
Derived series (no new sourcing, produced in the same build): GDP per capita, a growth decomposition (how much of growth came from population vs output per person - a log decomposition), and the annual growth rate + doubling time per period. Full source list and method: docs/layer_history_methodology.md in the project repository.
| Series | Source | Note |
|---|---|---|
| Population | HYDE 3.3 · McEvedy & Jones 1978 / Biraben 1980 · UN WPP 2024 | HYDE for ≤8000 BCE and ≥1 CE; monotone consensus 6000-2000 BCE; WPP for ≥1950 |
| Total world GDP | Maddison Project Database 2023 / OWID · Layer 1 ICIO (2025 anchor, $118.18T) | real-growth index × today's 118 T$ |
| Derived ratios | ratio of the two series above | per capita, growth decomposition, growth rate |
| Goods / services GDP | Layer 1 Markov trace; tax bridge distributed pro-rata into the two (site headline: goods $30.49T · services $87.69T · services share 74.2%). Earlier: structural-transformation literature (Herrendorf et al. 2014, Broadberry et al.) | gsyih_hizmet_pay - hand-entered, sourced estimate; no I-O table before 1995. The mechanism diagram above is the only place the tax is shown separately. |
| Surplus (net profit) | Layer 2 material-adjusted P&L, "true net profit" = GDP - wages (D1) - depreciation (P51C) - taxes (D2X3) - resource value | surplus_pay_gsyih - hand-entered, from the period shares of the four items; 2025 anchor 0.1905, pre-industrial 0.69 (Clark 2010, Piketty-Zucman 2014, ILO/PWT) |
7Perspective pages, The Extraction Trap, The Energy-Labour Multiplier
Two pages in the "Perspective" group of the Detailed Analysis section read Layer 1 with published headline statistics outside the main pipeline. Both are separate datasets: layer1_regional_manufacturing.json (The Extraction Trap) and layer1_energy_labour.json (The Energy-Labour Multiplier).
| Item | Source | Method |
|---|---|---|
| Regional manufacturing value-added (MVA) share | UNIDO, International Yearbook of Industrial Statistics 2024 · World Bank, World Development Indicators (manufacturing, value added) | regional share against world MVA $16T; extraction share from Layer 0 geography (§ layer0_methodology) |
| One-tonne iron → chassis value chain | Layer 0 (mine-mouth $110/t) · World Steel Association (crude steel) · last two steps illustrative | single-product chain; $/tonne step by step |
| World primary energy | Energy Institute, Statistical Review of World Energy 2024 (620 EJ ≈ 172,000 TWh) | primary energy ÷ world GDP = energy intensity (kWh/$) |
| Sectoral final-energy share | IEA, World Energy Balances 2024 | aggregated to 5 sector groups |
| Global employment | ILO, World Employment and Social Outlook 2024 (3.4 billion) | sectoral labour-hour intensity = employment share ÷ value-added share |
Sectoral kWh/$ and hours/$ are coarse (5 groups)
Energy intensity is derived from sector-energy ÷ sector-value-added, labour intensity from employment share ÷ value-added share. Real sectoral labour-hour intensity needs a hybrid input-output matrix such as EXIOBASE; this page gives 5 coarse groups.
MVA share can be derived directly
Regional MVA shares are taken from UNIDO; deriving them directly from the country value-added figures in layer1_country_view by regional aggregation is a future improvement. The last two steps of the value-chain example (processed sheet, car chassis) are illustrative, the order of magnitude is right, no decimal precision is claimed.
8Perspective pages, The Cost of Transformation + the workforce of the layers
The Cost of Transformation page added to the "Perspective" group, and the workforce strip that appears on every Overview, are compiled from published academic/institutional headline figures outside the main pipeline. Separate datasets: layer1_transformation_cost.json and layer_workforce.json (cross-layer).
| Item | Source | Method / note |
|---|---|---|
| Crude steel, 1.92 t CO₂/t · 21.27 GJ/t | World Steel Association, Sustainability Indicators 2024 (2003-2023 series) | Production-weighted average (BF-BOF, scrap-EAF, DRI-EAF); scope 1+2 + partial 3, 2023. The BF-BOF route alone is 2.32 t / 24.20 GJ. |
| Primary aluminium, 14,100 kWh/t · 14.8 t CO₂/t | International Aluminium Institute (2022/23), via JRC, JRC136525 (2024) | Energy: Hall-Héroult electrolysis electricity (AC, 2022); CO₂: cradle-to-gate (2023). About ¾ of emissions from the electrolysis step. |
| Cement clinker, 3.5 GJ/t · 0.83 t CO₂/t (0.53 calcination) | GCCA, Getting the Numbers Right, via JRC, JRC131246 (2023); calcination factor BREF 2013 | Thermal energy and gross CO₂ per tonne of clinker, world average (2019). Calcination CO₂ cannot be removed by switching fuels. |
| Primary copper, 22.2 GJ/t · 2.6 t CO₂/t · 70.4 m³ water · 190 t ore | Northey, Haque & Mudd 2013 (J. Clean. Prod. 40:118-128); grade Northey et al. 2014 (RCR 83:190-201) | Mine to cathode copper; producer sustainability reports, wide range (energy 10-70 GJ). Ore: 0.6% grade × 88% recovery. |
| Ammonia, 41 GJ/t · 2.4 t CO₂/t (best available tech 28 GJ) | IEA, Ammonia Technology Roadmap 2021 | Global average, all routes (coal-based production in China raises the average). About 80% of ammonia goes to nitrogen fertiliser. |
| Microchip (2 g, 32 MB DRAM), 630× the mass in secondary materials, 16,000× in water | Williams, Ayres & Heller 2002 (Env. Sci. Technol. 36:5504-5510) | Embodied-resource example, roughly year-2000 technology. Semiconductor-grade silicon takes about 160× the energy of metallurgical-grade. |
| Global exergy efficiency 10% (to useful work); global flow 475 EJ | Nakicenovic, Gilli & Kurz 1996 (Energy 21:223-237); flow Cullen & Allwood 2010 (Energy Policy 38:75-81) | Global exergy efficiency from fuel to useful work; the final-service efficiency is lower still. |
| Theoretical / practical energy reduction potential, 89% / 73% | Cullen & Allwood 2010 (Energy 35:2059-2069); Cullen, Allwood & Borgstein 2011 (EST 45:1711-1718) | Theoretical: efficiency limits of conversion devices. Practical: achievable redesign of building/vehicle/factory systems. |
| Global circularity rate 6.9% (2021) | Circle Economy, Circularity Gap Report 2025 | The 2018 report gave 9.1%; the metric was revised between editions (9.1→7.1). The decline direction is consistent. 106.5 Gt/yr processed material intake. |
| Employment by layer (L0 955M / L1 2,486M / L2 79M / L3 5M) | ILO modelled estimates, ILOSTAT (2024 update), by ISIC Rev.4 section, 2023; for L3 also SIFMA 2024 Capital Markets Fact Book + BLS NAICS 523/525 | L0 = A+B; L2 = K + real-estate estimate; L1 = every other section (industry plus all services). L0+L1+L2 partition the 3,520M employed; the L3 core is carved out of ISIC K, an estimate, band 3-7M. |
Transformation coefficients come from mixed system boundaries and years
The rows are from different system boundaries and vintages (steel 2023, aluminium 2022/23, cement 2019, copper 2008-11 reporting, ammonia 2018-20, microchip 1999), for orders of magnitude, not for summing or ranking to two significant figures. The copper figures come from voluntary corporate sustainability reports, with a self-selection toward larger, better-run operations.
Exergy "η ≈ 0.28" was re-scoped; the L1 workforce is a residual
The design mock's "η ≈ 0.28" is not a measured efficiency; it is the residual of the 73% practical saving potential (1 - 0.73). The page reports three separate, separately-sourced figures: measured exergy efficiency 10%, theoretical reduction 89%, practical 73%. On the workforce side: L0 is ISIC A+B and L2 is ISIC K+L, so L1 is everything else, all of industry and all of services, health, education and public administration included; the non-tangible services are entered as one derived line. Transport is reported by ILO bundled with information & communication; real estate (15-35M) and the L3 core (3-7M) are estimates. L0+L1+L2 partition total employment; L3 is carved out of L2.
9Country View, Domestic Processing Rate
The country map's third tab (added alongside production/consumption) shows how much of a country's own raw-material sector output (7 sectors: agriculture, forestry, fishing, coal, oil and gas, metal ores, other mining) stays within its own borders versus leaves as a raw export. In layer1_country_view.json, Domestic Processing Rate = 1 - raw exports / output; both are computed directly from the same ICIO row (that sector's sales to every country and every final-demand category).
| Country | Raw output | Raw exports | Domestic processing rate |
|---|---|---|---|
| China | 3.97 | 0.04 | 99% |
| United States | 1.51 | 0.27 | 82% |
| Turkey | 0.13 | 0.02 | 88% |
| Australia | 0.43 | 0.27 | 36% |
| Saudi Arabia | 0.36 | 0.23 | 37% |
| Norway | 0.25 | 0.22 | 13% |
A first-order measure, not a full value-chain trace
What happens to the "exported raw" share after it crosses a border isn't tracked - Norway's crude oil could be refined elsewhere and re-exported, this page doesn't see that. The only thing measured is how much of a country's OWN extracted/produced raw material stays inside its own borders and enters the next processing step there (its own industry or its own final consumption). A full global value-chain trace (Leontief inversion/Markov) is a much bigger undertaking, not done here.
Countries with negligible raw-material output are left as no data
Countries whose raw-material sector output is under $1 billion (Singapore, Hong Kong, Luxembourg, Cyprus, Malta, São Tomé and Príncipe) get no rate at all - with a denominator near zero, a percentage would be misleading; these countries stay uncoloured on the map.
The part of that output that is not consumed settles each year into deeds, factories and portfolios, it becomes wealth. Layer 2 · Wealth →