Wealth
The world's income statement and balance sheet, sources, method, validation
Think of the world as a single company: what would its income statement and balance sheet look like, where does each line come from, which year's data is it, how was it carried forward to today, and how was it verified.
Scope: the Income Statement (the world's GDP broken down into Wages/Taxes/Depreciation/Profit, 2022 base → scaled to 2025) and the Balance Sheet (the world's total Assets/Liabilities, by instrument and sector, largely real 2025 data). Both are a snapshot of a MOMENT (a stock); the trading volume (flow/turnover) that occurs in the world's financial markets during the year, which complements the balance sheet's STOCK view, is the subject of Layer 3.
1Income statement: line-by-line sources
| Line | Source | Raw data year |
|---|---|---|
| Net Sales (= World GDP) | Layer 1's ICIO-based world GDP estimate (layer1_resource_manifest.json, scaling.base_total_trillion_usd) - not from OECD, an independent figure derived from Layer 1's own 85-country × 50-sector ICIO matrix | 2022 → scaled to 2025 with the IMF ratio |
| COGS (Materials) - material-adjusted table only | Layer 0's at-source value column in layer1_resource.csv (40 commodities × volume × international reference price) | Weighted 2024-2025 |
| Wages (D1) | OECD SDMX DSD_NAMAIN10@DF_TABLE1_INCOME, applying the GDP-weighted ratio of 56 countries (55 OECD + China/NBS) to Layer 1's world GDP | 2022 |
| Taxes (D2X3, Net Production Taxes) | Same source, same method | 2022 |
| Depreciation (CFC, P51C) | OECD SDMX DSD_NAMAIN10@DF_TABLE2, GDP-weighted ratio of 40 countries applied to the world | 2022 |
2Reconciling the year mismatch, and two philosophies
World GDP and the Wages/Taxes/Depreciation figures are for the 2022 base year (the most recent year shared by OECD and ICIO). Layer 0's materials figure, however, is calculated from current commodity prices/volumes, so it already represents a magnitude around 2024-2025. The fix: apply Layer 1's own 2022→2025 scale factor (scaling.scale_factor) in REVERSE to shrink materials down to their 2022-equivalent:
Materials (2022) = Materials (current, Layer 0) / scale_factor
= $6.4851T / 1.172249 = $5.5322TIn the 2025 column, Layer 0's figure is used as-is. This is an approximation - the scale_factor is derived from the world's overall nominal GDP growth, and commodity prices (especially oil/gas) can be more volatile than GDP; if a materials-specific 2022 deflator is found, this line should be updated.
layer2_pnl_waterfall_world.json - standard SNA accounting. By definition, Value Added splits into 3 components: Wages + Gross Operating Surplus (Depreciation+Profit) + Net Production Taxes. Because intermediate-consumption cost is close to zero in the A/B (raw-material) sectors (nature doesn't send an invoice), almost all of the value added in these sectors appears as "profit."
layer2_pnl_waterfall_material_adjusted_world.json - an alternative framing: the portion of the A/B sectors' Operating Surplus that represents the market value of raw material extracted from nature is explicitly moved to a "Raw Materials" line. A deliberate departure from SNA that doesn't change GDP - it just redraws the boundary between "profit" and "materials."
Current presentation (2025, in classic corporate P&L order):
| Line item | SNA code | $2025 |
|---|---|---|
| Net Sales (= GDP) | B1GQ | $118.18T |
| - Raw Materials | (Layer 0 at-source value) | -$6.49T |
| = Gross Profit | $111.70T | |
| - Wages | D1 | -$57.93T |
| = EBITDA | $53.77T | |
| - Depreciation | P51C | -$21.23T |
| = Pre-Tax Profit | $32.54T | |
| - Taxes | D2X3 | -$10.03T |
| = Net Profit | $22.51T |
The difference between Net Profit without the material adjustment ($28.999T) and the adjusted Net Profit ($22.51T) is exactly equal to Layer 0's materials figure - the identity chain reconciles to the penny.
Source script: build_layer2_pnl_material_adjusted.py
3Balance sheet: what it is built from
How much is the world's total Assets and Liabilities; how much does each sector (Households, Non-Financial Corporations, Government, Financial Corporations) hold; in which instrument type (currency/deposits, loans, bonds, equity, insurance, real estate)?
| Source | Coverage | Freshness | Role in this balance sheet |
|---|---|---|---|
| OECD SDMX DSD_NASEC20@DF_T720R_A (non-consolidated) | 34-39 countries; Households, Government, Non-Financial Corporations; F1-F8 instruments | 2025 actual; missing country×cell combinations carried forward from 2024 using the country's own GDP growth | Main financial asset/liability skeleton |
| OECD SDMX DSD_NASEC20@DF_T710R_A (consolidated) | 33 countries; Financial Corporations | 2025 actual | Financial sector - the consolidated table was used to eliminate intra-sector cross-positions (using non-consolidated would double-count inter-financial-institution positions) |
| US Federal Reserve, Financial Accounts (Z.1) | US Financial Sector total + S12/F2 and S12/F6 instrument detail | 2025 actual | The OECD tables don't include a US Financial Sector - added directly |
| Swiss National Bank (SNB) | Switzerland Financial Sector total | 2025-Q4 actual | Same gap |
| Bank of Japan (BOJ, Flow of Funds) | Japan Financial Institutions total | 2025-Q4 actual | Same gap |
| PBoC, Sources & Uses of Credit Funds of Financial Institutions | China - Household/Corporate deposits+loans, Financial Sector total deposits+loans | 2025-12 actual | China isn't an OECD member, not in the table at all |
| PBoC, Financial Assets and Liabilities Statement (资金存量表) | China - 4 sectors × all instruments, FULL SNA financial balance sheet | 2024-12 actual (this table is published with a 2-year lag - the most recent available) | Filled China's Bonds/Equity+Funds/Insurance/Reserves gaps |
| NIFD, CASS National Balance Sheet Research Center - Macro Leverage Ratio | China - sector debt/GDP ratios | 2025-Q4 actual | Government's entirely missing debt + Corporations' missing bond portion |
| CASS National Balance Sheet Center (republished in an independent third-party article) | China - 4-sector Real Assets | 2019 actual, not scaled forward | OECD's Real Assets table (Table9B) doesn't cover China |
| IMF WP/2023/154 (Lam & Moreno Badia) | China - Government's unlisted SOE equity ownership + NSSF | 2019, carried to 2025-equivalent using the GDP ratio | PBoC's official SNA account doesn't cover unlisted SOE equity - a known gap in China's statistics |
| BIS, Total Credit (WS_TC) | 12 countries outside the OECD sample - Household loan debt | 2025-Q4 actual | Raised GDP coverage from 78.5% to 92.7% |
| World Bank, Broad Money (%GDP) | 10 countries outside the OECD tables - Financial Sector deposit-liability proxy | Mostly 2025, a few 2020-2024 (carried to 2025 with their own GDP growth) | Closest conceptual match to SNA F2 |
| OECD SDMX DSD_NASEC10@DF_TABLE9B | 32-33 countries (varies by category) - 4 sectors × 10 real-asset sub-categories | 2023 actual | Main source for Real Assets (housing, other buildings, machinery, intellectual property, inventory, natural resources, etc.) |
| IMF WEO Datamapper (NGDPD) | 229 countries - GDP | 2025 actual (with projections through 2031) | Denominator for extrapolation + each country's own growth rate |
4Balance sheet method
- Real data is sought first for every (country, sector, type, instrument) cell. OECD's 2025 table is the primary source; if not there, the 2024 table is carried forward to a 2025-equivalent using that country's OWN 2024→2025 GDP growth rate (first from OECD's own income-approach table, otherwise from the IMF WEO). Which cells are real vs. scaled-forward is marked row-by-row in the Source column of the output table.
- Major economies entirely missing from the OECD tables are added directly. The US/Switzerland/Japan with their own central banks' Financial Sector totals; China, with each of the 5 different sources above (PBoC × 2 tables, NIFD, CASS, IMF) added separately to the specific (sector, type, instrument) cell it covers - a cell covered by one source is never overwritten by another.
- Remaining gaps are filled with GDP-weighted extrapolation. For a given (sector, type, instrument) cell, the total value of the countries that HAVE real data is divided by those same countries' total GDP to get a ratio, which is then applied to world GDP: world_value = world_GDP × (sample_value / sample_GDP). This ratio is calculated SEPARATELY for EVERY (sector, type, instrument) combination - "Government's bond debt" and "Households' equity assets" use entirely different country samples and ratios; there is no single "world average" coefficient.
- Real Assets are built with a separate logic. Since OECD's SINGLE aggregate "real assets" code (NN) only exists for 8 countries, 10 sub-categories are instead used (Housing, Other Buildings, Machinery and Equipment, Intellectual Property, Inventory, Valuables, Agricultural Land, Mineral/Energy Reserves, Other Natural Resources), EACH extrapolated separately in its own country sample and then summed. China (not in this OECD table at all) is added separately with CASS's real data - to avoid double-counting, China's GDP share is first removed from the extrapolation sample base, then China's real value is added to the total.
- Known outliers are excluded case by case with a separate economic rationale - there is no automatic threshold rule. Example: Mexico's "Government Mineral/Energy Reserves" value was 201% of its own GDP (more than double the 2nd-ranked country) - a single country was skewing the world estimate, so it was excluded. By contrast, a similar-looking "Land" outlier (Korea, 112% of GDP) was NOT excluded, because it was consistent with the known "land-scarce-economy" pattern (France 44%, Japan 41% are also high) and excluding it would have over-corrected, pushing the estimate below McKinsey's.
5What came out
The total of the world's 4 sectors, without a sector breakdown, by instrument type (build_layer2_world_balance_sheet.py, 2025):
| Line item | Assets | Liabilities |
|---|---|---|
| Real Assets | $671.4T | - |
| Currency and Deposits (F2) | $194.7T | $180.5T |
| Debt Securities (F3) | $165.0T | $177.5T |
| Loans (F4) | $225.6T | $211.8T |
| Equity and Investment Fund Shares (F5) | $416.6T | $384.2T |
| Insurance/Pensions (F6) | $97.1T | $103.4T |
| Gold/SDR (F1) | $14.0T | $1.5T |
| Derivatives (F7) | $57.6T | $50.8T |
| Other Accounts Receivable/Payable (F8) | $64.9T | $62.8T |
| TOTAL ASSETS | $1,906.9T | |
| TOTAL FINANCIAL LIABILITIES | $1,172.5T | |
| NET WEALTH (Assets - Financial Liabilities) | $734.39T |
The Real Assets line is produced by its own separate pipeline, build_layer2_balance_sheet_real_assets.py (see §4/step-4 above), and is read DIRECTLY from that script's manifest here - it isn't recalculated separately, which prevents the two implementations from silently diverging.
340 (country, sector, type) rows rest on real data in at least one cell across 52 countries; the remaining share of world GDP was extrapolated using the method in §4/step-3 above.
6How it was tested
- Two independent PBoC tables confirm each other. Household deposits were calculated both from the "Sources & Uses of Credit Funds" table (2025-12, $23.89T) and from an entirely separate "Financial Assets and Liabilities Statement" table (2024-12→2025 scaled forward, $25.10T) - the two independent sources agree within 5%.
- NIFD's ratio-derived figure cross-checks against PBoC's direct-balance-sheet figure. China's Government bond debt was calculated INDEPENDENTLY from NIFD's leverage-ratio report ($13.42T, 2025-Q4 actual) AND from PBoC's separate SNA balance-sheet table (2024→2025 scaled forward, $12.59T) - they agree within 6%.
- The world's total assets nearly match the figure McKinsey Global Institute itself reports. The total assets of $1,906.9T here align with McKinsey's "$1.8 quadrillion" figure from its "Global Balance Sheet 2026" report, despite the two being built with two entirely independent methods (here: 52-country direct data + GDP-weighted extrapolation; McKinsey's: 23-country direct total, using licensed CEIC data for China).
- A second, independent comparison was built by replicating McKinsey's own method exactly. A footnote in McKinsey's technical appendix explains that the report's "world" figures aren't actually based on the full world, but on the DIRECT total of just 23 named economies (70% of world GDP). This method was replicated exactly (the same 23 countries, no extrapolation at all) - this 23-country test provides a second calculation path INDEPENDENT of the main (full-world) method, and the direction of the result (coming out below McKinsey's figure, the opposite of the full-world test coming out above it) confirms the coverage difference was correctly understood.
- Sector totals preserve identity consistency. Independent of its source, the instrument total of every (country, sector, type) row sums to the top-level sector total (Financial Assets, Financial Liabilities) down to the penny - there is no intermediate rounding/adjustment step.
- Real Assets were separately cross-checked against two independent external sources. McKinsey's 2024 figure for world real assets ($620T) is the same order of magnitude as the 2025-equivalent figure here ($671.4T) - the difference is consistent with a year of growth plus China now being added separately. Savills (a global real-estate research firm that uses IMF/BIS/WFE/World Gold Council data) puts world real estate at $393.3T for early 2025; the Housing+Other Buildings+Land total here (ex-China, since China's CASS data doesn't provide a category breakdown - $481.0T) is the same order of magnitude, consistent with the coverage difference (this total also includes infrastructure and commercial buildings) and the known overestimation risk in the Land line described below.
7Where the limits are
China's Real Assets figure is 6 years old (2019). GDP-ratio scaling was deliberately NOT applied - it was separately shown (during the McKinsey comparison in §6 above) that using the world-GDP ratio as a proxy for real estate systematically produces inflated growth, and China's 2021-2024 real-estate market downturn (per NIFD's own report: mortgage growth negative for 11 consecutive quarters) makes this risk especially large. The raw 2019 value is kept as a choice that's probably a slight underestimate, but known to be less wrong than GDP-scaling it forward.
China's Government Equity (SOE ownership) value rests on a 2019 academic estimate. PBoC's official SNA balance sheet only covers listed equity - the government's unlisted SOE ownership value (a large item, captured by the IMF study, at around 68% of GDP) is a known gap in China's own official statistics.
34 cells (a small share of the 340 rows) are still carried forward from 2024 using a 1-year country-GDP growth rate - for country×instrument combinations where OECD hasn't yet published 2025 data.
Real Assets' base year is 2023 (OECD's most recent year with the broadest coverage; country coverage drops from 33 to 20 in 2024) - carried to 2025 using the world-GDP ratio; this method's general tendency to produce slightly inflated growth for real estate was documented in §6 above, and was NOT fully resolved (the window was only narrowed to 2 years).
The Land line carries a known, NOT-YET-RESOLVED sample bias. The US reports no data at all for the "Land" (N211N) category, and the remaining sample of 15 countries is weighted toward land-scarce/land-expensive economies (Japan, Korea, Germany, France, the UK) - which likely causes the world Land value to come out inflated. The McKinsey/Savills cross-checks in §6 confirm the order of magnitude of the TOTAL, but the Land line ITSELF is still an uncorrected estimate that should be treated with caution. Next target: finding an alternative land-value source for the US (e.g. the real-estate breakdown in the Fed's Financial Accounts).
Layer 2 outputs outside the balance sheet (the Income Approach, the Investment-Type table) are still on the 2022 base year - these use Layer 1's own 2022 ICIO anchor directly, and fell OUTSIDE the scope of this balance-sheet update to 2025. (Real Assets' own separate pipeline, build_layer2_balance_sheet_real_assets.py, IS INSIDE this scope - the figures in §5 above come from that script.)
8Historical series (Historical Development page)
Layer 2's main work is a snapshot of end-2025. A separate Historical Development page (/layer2/tarihsel) tracks humanity's total accumulated net wealth from 1 CE (Rome) to today - there is no source giving a single world wealth figure before Rome, so the BCE points are empty. Wealth(year) = β(year) × GDP(year); β is the wealth-to-income ratio (Piketty & Zucman, "Capital is Back"): around 6 across history, 3.5 between the two world wars through destruction and inflation, 6.2 today (world net worth 734 T$ ÷ 118 T$ GDP). GDP is the Layer 1 historical series.
Derived series (same build, no new sourcing): net wealth per capita, the β curve, wealth turnover (in-year market flow ÷ wealth, from 1900), the financial share of wealth, and a growth decomposition + rate. The financial share of wealth is the single hand-entered series (0 at 1 CE → 50% today; idea: Goldsmith Financial Interrelations Ratio + the Piketty-Zucman financial/real wealth split; no exact series, low confidence). Full method: docs/layer_history_methodology.md §5 and §7b in the project repository.
| Series | Source | Note |
|---|---|---|
| Population | HYDE 3.3 · UN WPP 2024 | shared with the Layer 1 historical series |
| World GDP | Maddison Project Database 2023 · IMF WEO | "2025 USD equivalent", Layer 1 historical series |
| β (wealth/GDP) | Piketty & Zucman, "Capital is Back" (2014) | 6 → 3.5 (interwar) → 6.2 (today) |
| Financial share of wealth | hand-entered, sourced educated guess | Goldsmith FIR + Piketty-Zucman; low confidence |
9Perspective pages, Concentration, Accumulation Rate
Two pages in the "Perspective" group of the Detailed Analysis section. Accumulation Rate (/layer2/detay/birikim-hizi) derives mostly from the historical series in §8 and Layer 2's own P&L / investment-type outputs (layer2_income_approach_world.json, layer2_investment_by_asset_world.json, 2022 base, as noted in §7). Concentration (/layer2/detay/yogunlasma) is a separate dataset (layer2_wealth_concentration.json) compiled from published wealth-distribution statistics, outside the main pipeline.
| Item | Source | Method |
|---|---|---|
| Top 1% / 10% / bottom 50% wealth share | Credit Suisse / UBS, Global Wealth Databook 2023 · WID.world 2024 | household net-wealth base; shares direct |
| Global wealth Gini | Credit Suisse, Global Wealth Databook 2023 | 0.88 |
| Wealth pyramid (4 bands: headcount + share) | Credit Suisse, Global Wealth Databook 2023 (5.4 billion adults) | threshold-based bands |
| Regional wealth share | UBS, Global Wealth Report 2024 | N. America / Asia-Pacific / Europe / LatAm-Africa-India; shown beside population share |
| Annual flow into accumulation | Layer 2 P&L (net profit + depreciation) · layer2_investment_by_asset_world.json | share of output; asset-type split is 2022 base, scaled to the current year |
Household wealth base ≠ Layer 2 net wealth
UBS/Credit Suisse measure $454T of household net wealth; Layer 2's net wealth ($734T) is a broader base (includes public + corporate). The Concentration page does not reconcile the two totals, it presents shares, the Gini and the pyramid structure, not the absolute total.
The top-1% share moves year to year
Rounded headline value: Credit Suisse gives 44.5% for end-2022, UBS 47.5% for 2023; the page shows 45%. Vintage is in the table above.
10Perspective pages, Intangible Capital
The Intangible Capital page added to the "Perspective" group (/layer2/detay/maddesiz-sermaye) reads the detachment of wealth from physical assets through published academic and institutional headline figures, a separate dataset (layer2_intangible_capital.json), outside the main pipeline.
| Item | Source | Method / note |
|---|---|---|
| S&P 500 intangible share (1975 17% → 2025 92%) | Ocean Tomo / J.S. Held, Intangible Asset Market Value (IAMV) Study | Market value - book (tangible) value = intangible. A single-vendor calculation; used as a direction indicator, not a precise level. |
| Intangibles are a large and rising share | Corrado, Hulten & Sichel 2009 (RIW 55:661-685); Corrado, Haskel, Jona-Lasinio & Iommi 2022 (JEP 36:3-28); Haskel & Westlake 2018; Peters & Taylor 2017 (JFE 123:251-272); Lev & Gu 2016 | Multiple independent peer-reviewed research programmes confirm the direction: intangible investment and capital stock are large and rising. |
| US business intangible investment $1,226 billion/yr (2000-03 avg.) | Corrado, Hulten & Sichel 2009 (RIW 55:661-685), Table 1 | CHS definition (software, R&D, brand, design, organisational capital). Same order of magnitude as tangible business investment. |
| US intellectual-property-products investment 5.5% of GDP (2024 Q1); 31.3% of private fixed investment | US BEA, NIPA private fixed investment; Richmond Fed Macro Minute (2 Jul 2024); BEA 2013 Comprehensive Revision | The narrow definition capitalised in the national accounts (software + R&D + entertainment originals). The 2013 revision raised the level of US GDP for 2012 by 3.6%. |
| 272,600 PCT patent applications in 2023; top 5 origins 78.3%; largest applicant (Huawei) 6,494 | WIPO, Patent Cooperation Treaty Yearly Review 2024 | Application counts direct; origin share is the sum of the top 5 countries. |
| Top 10 S&P 500 firms 38% of index value (2024; 19% in 2010); US = 73.9% of MSCI World | S&P Dow Jones Indices (year-end 2024); MSCI, MSCI World factsheet (31 Dec 2024) | Market-value concentration; a figure that moves year to year, vintage 2023-2024. |
The Ocean Tomo series is a single-vendor estimate, direction robust, exact number not
The entire 50-year S&P 500 intangible-share series is Ocean Tomo's own market-value-minus-book-value calculation (a consultancy, not peer-reviewed). Peer-reviewed work confirms the direction only (large and rising); the exact '90%' or '92%' is one vendor's estimate and is sensitive to how 'tangible/book value' is defined. Part of the rising share reflects buybacks and low interest rates inflating market value, not intangible accumulation alone.
The investment share is data-release-sensitive; a mock claim was dropped
As a share of GDP the intangible investment figure varies by data release (INTAN-Invest, EU KLEMS and national-accounts-based series give different levels); the direction (tangible falls, intangible rises) is consistent across all of them, the page gives two US anchors plus one qualitative direction row, with no clean global '% of value added' number. The design mock's '76% of critical patents in 100 firms' was dropped because no source stands up that framing; the WIPO PCT and S&P/MSCI concentration figures replace it.
This accumulated wealth does not stay still; the markets turn it over at many times its own size. Layer 3 · Markets →