Urdais Bitcoin Wealth Index (UBWI)

Version 1.2.0, 15 September 2026. Status: approved for production. Prepared under the Urdais methodology framework. UBWI is an estimate calibrated to observed economies, not a census of world wealth: roughly two fifths of its denominator is a disclosed, versioned model, and every surface carrying the value says so. This version amends the numerator's supply rule, replacing an externally reported circulating-supply dataset with issuance derived deterministically from Bitcoin's own block subsidy schedule. That removes the last external rights dependency from the numerator — by removal, not by clearance — and the publication gate passes for the first time, with no finding. The first UBWI value is published under this version. No gate threshold was changed to reach it.

Purpose and Scope

UBWI answers one question and no other.

What share of all presently existing global wealth is represented by Bitcoin?

$$\text{UBWI} = \frac{\text{Bitcoin Market Capitalization}}{\text{Total Global Wealth}} \times 100$$

The published statistic is a percentage of Total Global Wealth. It is not an index level, carries no base date and no base value, and is never expressed in points. Any presentation of UBWI as a points series is wrong and must be corrected.

What UBWI Is Not

UBWI is not a measure of Bitcoin holder concentration, address distribution, ownership dispersion, or wealth inequality among holders. It is not Bitcoin's share of money, of monetary aggregates, or of any "investable assets" universe. It is not a valuation model, a fair-value estimate, or a forecast. Each of those is a different economic object; none of them may be published under this name.

The Numerator: Bitcoin Market Capitalization

Definition

The Bitcoin market capitalization is issued Bitcoin supply multiplied by a reference price in United States dollars, both observed at a single instant.

$$\text{BTC market cap} = S_t \times P_t$$

Supply

$S_t$ is the protocol-scheduled cumulative block subsidy through the reference block height at time $t$, expressed in BTC to eight decimal places, together with the block height it was taken at. The block height is recorded on every observation and is the reproducibility key: a supply figure without its height cannot be checked.

Urdais derives Bitcoin supply deterministically from the Bitcoin block subsidy schedule through the reference block height. This represents protocol-scheduled issuance rather than an externally reported circulating-supply dataset.

The derivation

The subsidy schedule is consensus arithmetic, not a dataset:

genesis is block height 0;
the initial subsidy is 50 BTC, i.e. 5_000_000_000 satoshis;
the subsidy halves every 210_000 blocks, by integer division, which is the protocol's own rule rather than an approximation of it;
transaction fees are not included: fees are redistributed, not issued, and counting them would double-count coins already counted in an earlier block's subsidy;
the subsidy truncates to zero from halving era 33 — block height 6_930_000 — because 5_000_000_000 lies between $2^{32}$ and $2^{33}$, so the integer halving series reaches one satoshi at era 32 and zero at era 33.

Height convention: the reference height is inclusive. $S_t$ is the cumulative scheduled subsidy for all blocks up to and including the reference height. Cumulative supply at height 0 is therefore 50 BTC — one block's subsidy — and not zero. This is stated explicitly because an off-by-one here is a 3.125 BTC error in the current era that no downstream check would catch.

Integer arithmetic. All accumulation is in integer satoshis. A real supply is on the order of $2 \times 10^{15}$ satoshis, within a factor of five of the largest exactly representable integer in IEEE-754 double precision, so a floating-point accumulator would be silently wrong rather than loudly wrong. Conversion to decimal BTC happens once, at the display and market-capitalization boundary.

Invariants. Supply is non-negative, monotonically non-decreasing in height, and never exceeds the 2_100_000_000_000_000-satoshi cap. The schedule in fact asymptotes to 20,999,999.9769 BTC and never reaches 21,000,000, because integer truncation discards a remainder at every halving.

What this quantity is not

It is not exact circulating supply, spendable supply or recoverable supply, and the published surface never calls it any of those.

Miners have historically, on occasion, underclaimed coinbase rewards — claiming less than the schedule permitted — and those coins were never created. Protocol-scheduled issuance is therefore very slightly above exact created supply. Urdais's own Phase 1 comparison put the difference at approximately 44 BTC, but that figure was derived by differencing the schedule against a supply reading from blockchain.info, the very source this version retires; it is an estimate carried over from a retired source and is not independently supported by any artifact Urdais retains. It is recorded here as an order of magnitude and nothing more. At roughly 0.0002 % of supply it is immaterial at UBWI's published precision, which is the substantive point and does not depend on the exact figure.

The genesis coinbase is famously unspendable and is nonetheless included, because excluding it would be a spendability adjustment and this version makes none.

Why the external supply source was retired

Versions 1.0.0 and 1.1.0 read $S_t$ as claimed issuance from Blockchain.com's Explorer API. Blockchain.com's retained terms grant retrieval and scope the Explorer "solely for informational purposes"; they do not grant the commercial derived-index publication Urdais performs. That was the last external rights blocker on the numerator, and under 1.1.0 the publication gate refused on exactly that ground.

Version 1.2.0 does not clear those terms and does not waive them. It removes the dependency: the quantity is now computed from the block height, and no supply figure is retrieved from Blockchain.com or from anyone else. The Blockchain.com terms artifact is retained as the evidence for the retirement, not as a licence Urdais relies on for any published quantity.

The trade is deliberate and it is a change of construction, versioned as such: claimed issuance is closer to coins actually created, while scheduled issuance is reproducible by anyone from the height alone and depends on no third party's authority. At a difference on the order of 0.0002 % of supply, reproducibility is worth more to a published index than a correction smaller than its own rounding.

Rights state

The supply quantity carries the rights basis derived_from_protocol. It is deliberately not cleared, not expressly granted and not inferred_permitted: no third-party data licence is relied upon for the quantity, so no statement about a third party's permission is made in either direction.

Reference block height

The height itself is still observed, and it is treated as a network fact rather than a proprietary dataset: it is the same integer for every honest observer of the chain. It is read from two independent sources — mempool.space/api/blocks/tip/height and blockchain.info/q/getblockcount — and every observation freezes each source's identity, its raw returned value, its retrieval timestamp and its provenance.

The production rule is exact agreement and fail-closed. Both readings must be valid non-negative integers and must agree exactly. Heights are never averaged, and the higher or lower reading is never taken silently: a tolerance on a block height would be an invented licence to publish a supply the chain never scheduled. On disagreement the gate refuses and the operator re-reads the tip.

Reading blockchain.info/q/getblockcount for one consensus integer is a different act from taking a supply quantity on Blockchain.com's authority, and Section 20 of its retained terms grants the access and use that cross-verification requires. The height sources establish the integer and nothing else; they are not the source of the supply value.

Lost coins

No lost-coin adjustment is made. Supply is used unadjusted.

This is a decision, not an oversight. Published estimates of lost Bitcoin span roughly 2.3 million to 5.6 million BTC — a range wider than a sixth of supply — and every one of them reduces to a choice of dormancy threshold (five years, seven years, ten years, or "inactive since July 2010") for which no source states a derivation. Dormant coins are not provably lost: fork and airdrop events have reawakened long-dormant outputs. The only deterministic component, provably unspendable output (unclaimed coinbase, burn addresses, OP_RETURN), is small enough to be immaterial here and is not available under a redistributable licence.

A free-float or lost-coin-adjusted numerator would therefore embed an unfalsifiable assumption in a series whose whole purpose is to be inspectable. If a lost-coin estimate ever acquires a documented, deterministic, scheduled and reproducible construction, admitting it becomes a versioned methodology change and does not restate history.

Price

$P_t$ is the latest valid Chainlink BTC/USD Data Feed reference price on Ethereum mainnet available at the UBWI calculation timestamp.

This replaces the three-venue exchange median that version 1.0.0 specified. The change is an amendment to the price rule, not a substitution of one venue for another inside an unchanged rule, and it is versioned as such.

Why the three-venue median was retired

Version 1.0.0 reasoned that reading public venue tickers directly removes the licensing dependency a vendor aggregate would carry. That reasoning was correct about vendors and was never a statement about the venues. When the venues' own terms were retrieved and read, three of the four numerator interfaces did not permit the use Urdais actually makes of them — retain the reading, compute a commercial index from it, and display both — and the fourth granted exactly that use only under a signed agreement Urdais does not hold:

Coinbase prohibits collecting, caching, aggregating and storing the data, prohibits sharing it with third parties without prior written authorization, and prohibits recording it by automated program at all.
Kraken grants use of its content "only for your own benefit" and lists commercial exploitation and making it available to a third party among the prohibited acts.
Bitstamp permits incorporation, redistribution and derived calculations commercially — to a company that has signed its Data License Agreement. Urdais has not signed one and did not seek one.

Reproducing a construction yourself does not reproduce the permission to publish it. Those findings, and the retained terms artifacts behind them, are preserved in the source-rights record; retiring the rule does not retract them, and the retired observation itself is preserved in the codebase rather than deleted.

The feed

Network: Ethereum mainnet, chain ID 1.
Proxy: 0xF4030086522a5bEEa4988F8cA5B36dbC97BeE88c.
Pair, from description() on the proxy: BTC / USD.
decimals(): 8.
Heartbeat: 3600 seconds.
Deviation threshold: 0.5 %.
Product: a standard push-based Data Feed reference price, BTC/USD-RefPrice-DF-Ethereum-001.

Every value above was read from the live contract through two independent public RPC endpoints, which returned byte-identical rounds, and the heartbeat and deviation threshold were read from Chainlink's own feed metadata document — the file docs.chain.link itself renders.

The feed is not Data Streams and not Smart Value Recapture, which are different Chainlink products with different delivery channels and different terms.

The proxy is the methodology pin. Reads go through latestRoundData() on the proxy and never through a transient aggregator implementation. Chainlink upgrades the aggregator behind a stable proxy by design, and an ordinary aggregator upgrade behind the approved proxy is not a methodology change: the aggregator address, its typeAndVersion and the round's phase are frozen as observation lineage on every published point, so any round stays auditable after an upgrade. A change to the proxy, the pair, the network, the product type or the source class fails closed pending methodology review.

Validity

A price observation is valid only if every applicable check passes. There is no partial acceptance and no fallback.

Network. Chain ID must be 1.
Feed identity. The proxy must match the approved address, description() must equal BTC / USD, and decimals() must equal 8.
Round. latestRoundData() must succeed with answer > 0 and updatedAt > 0. The round id's decomposition must agree with the frozen phase and aggregator round. answeredInRound is frozen but never read as a validity condition: Chainlink's own API reference marks it deprecated, and the old answeredInRound >= roundId idiom is a test on a field that no longer carries that meaning.
Staleness. retrieval_timestamp - updatedAt <= 3600 seconds, the documented heartbeat. No grace period is added. Chainlink's own documentation directs consumers to "use the updatedAt value from the latestRoundData() function to make sure that the latest answer is recent enough". Beyond 3600 seconds the numerator is classified stale and UBWI is withheld.
RPC. If the endpoint is unavailable, times out, returns the wrong chain id, disagrees with a second endpoint, or the block cannot be re-fetched, the observation fails closed and nothing is published.
Deprecation. If the proxy reverts, returns zero, or the feed identity no longer matches, the observation fails closed. Urdais does not silently switch networks or feeds.

Lineage frozen on every published point

Chain id; proxy address; aggregator address and its type and version where determinable; proxy round id; phase id; aggregator round id; the raw answer; decimals; the normalized USD value; startedAt; updatedAt; the retrieval timestamp; the block number and block hash the read was pinned to; the RPC endpoint's identity (never a secret, and no endpoint Urdais uses carries one); the Chainlink rights state; and the methodology version.

What a Chainlink reference price is, and is not

It is not a spot exchange price. Chainlink's own documentation says Data Feeds "aggregate many data sources and publish them onchain". Urdais cannot independently reconstruct the underlying source basket, and the feed does not disclose it. This is a real and stated limitation of the numerator.
It is not streaming. The feed updates when the price deviates past the 0.5 % threshold or when the heartbeat elapses, so an observation may sit up to 0.5 % away from the wider market between updates, and up to an hour old. That is the numerator's price uncertainty, and it remains roughly two orders of magnitude smaller than the denominator's sensitivity range.

A directly published market-capitalization series is still not used. The surveyed vendors either do not disclose how their price is formed, or forbid the use Urdais would make of it, or both. Where a vendor series is retained, it is retained as a comparison diagnostic and never as the numerator.

Update frequency

The numerator is observed at least daily and may be observed more often. Each observation carries its own instant. The numerator is the only part of UBWI that moves between denominator vintages.

The Denominator: Total Global Wealth

"Total Global Wealth" is the formal term. "Global wealth supply" is not used anywhere in Urdais.

Definition

Total Global Wealth is the consolidated net worth of the world economy: the market value of all non-financial assets owned by resident institutional units of every economy, plus monetary gold bullion, plus crypto assets that carry no corresponding liability, less nothing further — because at world scale every remaining financial asset is another unit's liability and the two cancel.

This is the national-accounts definition of national wealth, summed to the world. The 2008 System of National Accounts states it at paragraph 13.4:

"Net worth is defined as the value of all the assets owned by an institutional unit or sector less the value of all its outstanding liabilities. For the economy as a whole, the balance sheet shows the sum of non-financial assets and net claims on the rest of the world. This sum is often referred to as national wealth."

The 2025 SNA carries the same sentence at paragraph 14.4, with "national wealth or national net worth".

Summing that identity over every economy removes the second term, because one economy's claim on the rest of the world is another economy's liability to it. What survives is the world's stock of non-financial assets. The World Inequality Database publishes exactly this identity for its national wealth series: "[Market-value national wealth] = [National non-financial assets] + [Net foreign assets]".

The anti-double-counting rule

Asset-class market values are never summed. This rule is the reason the definition is written the way it is, and it is stated here so that any future component list can be checked against it.

Four specific prohibitions follow.

The consequence is large, not marginal. A published global balance sheet for end-2024 puts the world's gross assets at $1.7 quadrillion and its consolidated net worth at $600 trillion: consolidation removes a factor of about 2.8. A denominator built by addition would be wrong by that order.

Debt claims are not wealth. A bond, a loan, a deposit and a banknote are assets of the holder and liabilities of the issuer. Adding the world's debt securities outstanding to the world's real assets counts the financed asset twice. Global broad money is a liability of the banking system and is excluded for the same reason.

Corporate equity is not added to corporate assets. A listed company's market capitalization is a claim on the same factories, inventory, cash and intellectual property that appear in the non-financial asset stock. Counting both duplicates the company. In the market-value national-wealth construction the duplication is resolved in the other direction from the one people expect: corporate net worth is consolidated into the sectors that hold the equity, so household and government wealth carry the market value of the businesses, and the businesses are not counted again.

Household-net-worth frameworks and national-balance-sheet frameworks are not mixed. A household-sector measure excludes government assets and government debt by construction; a national measure includes both. Adding a household total to a government total taken from a different framework produces neither, and adding either to an asset-class aggregate produces a number with no accounting meaning. A denominator is taken from one framework at a time.

A component that overlaps another component is not additive. Global real-estate market value, produced capital, natural capital and agricultural land all overlap each other under different valuation bases. Where UBWI displays such figures they are displayed as cross-checks, explicitly marked non-additive, and they never enter the denominator sum.

Inclusions

Within the definition, and therefore inside the denominator whenever the chosen source measures them:

residential and commercial real estate, at market value, including the land beneath it;
produced capital: machinery, equipment, structures, infrastructure, inventories, and intellectual property products recognised as assets;
land, subsoil assets, timber, fisheries and other natural capital within the SNA asset boundary, that is, natural resources over which ownership rights are enforced and from which economic benefits accrue;
valuables: precious metals held as a store of value, gemstones, art and antiques. Monetary gold bullion is included as the single financial asset with no counterpart liability; non-monetary gold and silver are valuables;
businesses, public and private, through the consolidated treatment above, never as an equity market capitalization added on top of the assets;
crypto assets without a corresponding liability, which the 2025 SNA places inside the asset boundary as non-produced non-financial assets (paragraph 4.117), on the reasoning that miners "are considered to be producers of validation services, not as producers of the assets themselves". Bitcoin is such an asset.

Exclusions

Human capital is excluded. It is the discounted value of expected future labour income, not a presently held asset, and its valuation is dominated by assumptions about discount rates, earnings growth and horizons. The SNA excludes it from the asset boundary for the same reason: "Human capital is also not treated by the SNA/BPM as an asset in the integrated framework of national accounts" (2025 SNA 4.121). Where a source publishes a comprehensive-wealth total that includes human capital, the human-capital component is stripped out and the stripping is shown.

Also excluded:

consumer durables, which the SNA holds outside the asset boundary and reports only as a memorandum item;
gross financial claims, for the reason given above;
natural resources of no economic value, and environmental assets over which no ownership rights are enforced;
contingent and constructive liabilities, and unfunded state pension entitlements, consistently with the SNA treatment. Funded pension assets are included through the sector that holds them.

Government and public assets

Government assets and government liabilities are inside the denominator. Total Global Wealth is a national measure, not a private one: public infrastructure, public land and state-owned enterprise capital are presently existing wealth, and government debt is a liability of the public sector and an asset of whoever holds it, so it cancels globally exactly as private debt does. Net public wealth, positive or negative, is published as a denominator component.

This is the principal reason a household-net-worth source cannot be used unmodified as the denominator: it excludes state assets and state debts by design.

Bitcoin inside its own denominator

Because the 2025 SNA places Bitcoin inside the asset boundary, Bitcoin is part of Total Global Wealth, and UBWI is a genuine share rather than a ratio to something Bitcoin sits outside of. The denominator is therefore

$$\text{TGW}_t = W_v + \text{BTC market cap}_t$$

where $W_v$ is the published world-wealth vintage $v$, which — because every currently published vintage predates national implementation of the 2025 SNA — contains no crypto assets. UBWI is bounded in $[0, 100]$ by construction.

Adding Bitcoin lowers UBWI by about a quarter of one percent of its own value at present magnitudes. The adjustment is immaterial and is made anyway, because the alternative is a statistic that claims to be a share of a whole that excludes its own numerator.

Crypto assets other than Bitcoin are not added to the denominator. They belong there under the same paragraph, they are not measured here, and their omission biases UBWI upward. This is a named limitation, quantified with every published value once a redistributable measure of non-Bitcoin crypto capitalization exists.

Production Denominator Construction

Total Global Wealth is constructed, not read. No compiler publishes a world national-wealth total at market value and market exchange rates under terms that permit redistribution, and the research that established this is recorded in the Phase 1 to Phase 2C source studies. UBWI therefore builds its denominator as

$$\text{Total Global Wealth} = \text{Observed Rights-Cleared Wealth} + \text{Modeled Residual Wealth} + \text{Bitcoin Market Capitalization}$$

Observed wealth

The observed leg is the sum of national balance sheets for the economies Urdais can both observe and publish. Each component carries its compiler's own series identifier, its own reference date, its own currency and FX conversion, its land and consumer-durables treatment, and its rights state. Components are summed, never averaged, and no component is bridged, interpolated or rolled forward.

Observed does not mean harmonised. Two harmonisation layers over the same national compiler can disagree materially — France's net foreign position differs by EUR 379 bn between the OECD and Eurostat — so a component records the interface it actually came from, and a disagreeing route is retained as superseded rather than silently discarded.

Taiwan, admitted in version 1.1.0

Taiwan is the eighteenth observed constituent and the only one added since version 1.0.0. It is what brings the modelled share of the denominator under the 40 % ceiling, and it clears that ceiling by 0.24 percentage points — narrowly enough that a single revision could un-clear it, which is worth stating rather than presenting the bound as comfortably met.

The concept. The Directorate-General of Budget, Accounting and Statistics compiles National Wealth Statistics on a 2008 SNA basis, publishing a year-end total-economy net worth. For 2024, released 29 April 2026: NT$2,589,685 × 10⁸, which is net non-financial assets NT$2,080,329 × 10⁸ plus net financial assets NT$509,356 × 10⁸. The accounting identity reproduces exactly, and DGBAS's own Table 1 note 1 states that net financial assets equal net foreign financial assets, because domestic claims and obligations offset within the national economy — the same construction as the Eurostat components.

Consumer durables are stripped, as for every other component that includes them. DGBAS reports household durable and semi-durable properties (NT$27,885 × 10⁸) and household cars and motorcycles (NT$31,921 × 10⁸) separately, together NT$59,806 × 10⁸. The accepted figure is therefore NT$2,529,879 × 10⁸, converted at the Central Bank of the Republic of China's interbank spot market closing rate for 31 December 2024 of NT$32.781 per US dollar — matched to the component's own reference date, as the FX policy requires. DGBAS's national-accounts workbook publishes a 32.11 rate for 2024 but labels it "Average of daily figures", which is a period average and the wrong basis for a year-end stock. The accepted contribution is US$7.7175 trillion.

The figures are taken from DGBAS Table 4, which publishes in units of 100 million NT$, rather than from Table 1, which publishes the same totals rounded to two decimal places of NT$ trillions.

Three caveats travel with Taiwan and none is hidden.

Land is valued at 公告現值, the announced current land value — an administrative assessment, not a market price. DGBAS states this in Table 1 note 3. Land is included in full (NT$134.12 trillion, 52 % of net worth), but on an administrative basis that almost certainly understates it: DGBAS's own alternative series, re-valuing residential, industrial and commercial land at market price, raises net worth by NT$10,119 × 10⁸, or 0.39 %. Urdais takes the announced-value headline, so Taiwan's contribution is if anything conservative.
Taiwan is absent from the World Bank's Changing Wealth of Nations cross section. It contributes wealth to the denominator while contributing nothing to the residual model's calibration weights, so the reported observed share of CWON wealth understates the observed set from this version onward. The calibration factor itself is unaffected: Taiwan is in neither side of the ratio.
Taiwan is not a World Development Indicators country, so its GDP weight comes from DGBAS's own nominal GDP in US dollars (US$801.529 bn for 2024) rather than from the same source as every other component's. This is stated because it is a basis inconsistency, small and unavoidable.

Double counting is avoided, and the reason is a retrieved World Bank statement rather than an assumption. The World Bank states that "Taiwan, China is not listed as a separate country for World Development Indicators. For most indicators, Taiwan, China is not included in the estimate for China, but is added to the world aggregate and the aggregate for high-income countries." The world GDP denominator therefore already includes Taiwan, and Taiwan does not sit inside China's figure. Adding Taiwan to the observed set correctly removes its GDP from the unobserved remainder the residual model values, and cannot double count it against China, which remains disclosed as unobserved.

Rights. DGBAS releases everything its sites publish under the Open Government Data License, Taiwan 1.0 — free of charge, non-exclusive, sublicensable, unlimited in time and territory, covering reproduction, adaptation, editing, public transmission and the development of derivative products, and stated in terms to be not afterwards withdrawn and to need no separate written permission. Attribution is the one condition and it is a real one: the licence text provides that a user who fails to attribute is treated as never having been granted the rights. The Central Bank of the Republic of China declares the same licence for its exchange-rate data. Both declarations are retained and hashed.

Modeled residual wealth, and why modelling is structurally necessary

The economies Urdais does not observe are not a research backlog. They are economies that do not compile the thing being measured. China does not publish a national balance sheet with a market-valued net-worth total; neither does India. No permission, retrieval or budget creates a statistic a national statistical office does not produce. Phase 2C measured the feasible frontier — the rights-cleared coverage reachable from balance sheets that exist today — at 55.72 % of world GDP, and a counterfactual ceiling of 62.93 % that would require nineteen further statistical offices to begin valuing land.

A denominator that included only observed economies would not be world wealth; it would be the wealth of the rich world, presented as the world's. Modelling the remainder is therefore the honest option, provided the model is versioned, disclosed, and never called an observation.

The rule. Unobserved world GDP at market exchange rates is valued at the observed set's own wealth-to-GDP ratio, scaled by a calibration factor measuring how much poorer per unit of output the unobserved world is in the one cross section where both are measured comparably — the World Bank's Changing Wealth of Nations, 2020:

$$\text{Modeled Residual Wealth} = k \cdot R \cdot \text{Unobserved World GDP}$$

where $R$ is the observed-set wealth-to-GDP ratio computed from the components themselves and $k$ is the CWON tail calibration. The rule is deliberately simple. Phase 2A backtested regional and income-group refinements and found they do not reliably beat a single world ratio — regional grouping is worse at six of eleven cut-points — so the model does not invite elaborations that add opacity without accuracy.

The imputation rule is versioned reference data, not application code, so a value computed under one rule stays explicable after the rule changes. History is never restated.

The observed and modelled shares of Total Global Wealth are published beside the value, in those words. Modelled wealth is never described as observed, and the headline is never shown without its modelled share and its sensitivity range.

Accounting Framework

Four candidate frameworks were compared. The comparison is recorded in the Phase 1 source study; the conclusion is stated here because it determines everything above.

National balance sheets on the SNA basis are the chosen framework. They are the only framework in which the consolidation is an accounting identity rather than an adjustment: the double-counting is removed by construction, not by netting one estimate against another. They cover every sector, so government wealth is included. They value assets at market where market prices exist. And the residual they leave — the world's measured net foreign position, which should be zero and is not — is a published, quantified statistical discrepancy rather than a hidden one.

Household net worth is rejected as the primary framework. It is well measured and internally consistent, and it is the basis of the most widely cited global wealth figures, but it excludes government assets and debts by definition and so cannot answer a question about all presently existing wealth. It is retained as a cross-check.

Comprehensive wealth frameworks — produced capital plus natural capital plus human capital plus net foreign assets — are rejected as the primary framework for two reasons: human capital dominates them, so the measure is mostly the thing UBWI must exclude; and their produced capital is a perpetual-inventory replacement cost and their natural capital a discounted resource rent, neither of which is a market value. They are retained as a structural cross-check on composition, and as a floor.

Asset-class aggregation — adding real estate to equities to bonds to gold — is rejected outright. It is the double-counting error the anti-double-counting rule exists to prevent, and no weighting of the components repairs it.

Where unavoidable uncertainty remains

Four sources of uncertainty survive the framework choice and cannot be assumed away.

The world does not balance. Summing every economy's foreign assets and foreign liabilities should give zero. Measured, it gives a negative number on the order of one percent of world wealth. That residual is the accumulated effect of unrecorded cross-border positions, offshore holdings, and valuation and coverage differences between national compilers.

Coverage is incomplete and uneven. No compiler publishes market-value national balance sheets for every economy. The economies that do publish them are disproportionately high income, and the rest are modelled.

Valuation basis is not uniform. Dwellings and land are marked to market where a property price index exists and are otherwise modelled. Unlisted business equity is valued by analogy to listed equity in some national accounts and at book value in others. Natural capital is a discounted rent, not a price.

Purchasing-power and market exchange rates give materially different totals. The conversion applied to reach a market-exchange-rate denominator is itself a source of error, and its direction is known: see the vintage section.

Foreign Exchange Policy

End-period FX for end-period stocks. A national balance sheet is a stock at an instant, and it is converted at the last quoted reference-rate fixing at or before that component's own reference date. Phase 2B measured the error from the wrong basis at up to 6.6 % for a single country, which is larger than several economies' entire contribution to the denominator.

The fixing date is per component, not per vintage. Australia's national balance sheet is as at 30 June; every other economy in the observed set is a calendar year-end. A single harmonised fixing date would either treat a 30 June stock as a 31 December stock or select the wrong fixing for it.

One FX source per component, rights-cleared, recorded. Every non-USD component except Taiwan is converted at the European Central Bank's euro foreign exchange reference rates, whose terms permit free use with accurate reproduction and citation of the ECB, and which require that any modification be stated explicitly. Converting a stock to USD is such a modification and is disclosed on every component's FX lineage.

Taiwan is the one exception and it is named rather than absorbed: the ECB publishes no New Taiwan dollar reference rate, so Taiwan's stock is converted at the Central Bank of the Republic of China's own interbank spot market closing rate for the component's reference date, under the same OGDL-Taiwan 1.0 licence as the stock itself. A second FX source is a thing to notice; it is recorded on the component, and a source is never switched without recording the change.

Denominator Vintage Policy

Comprehensive world-wealth estimates are annual, refer to a year-end, and are published six to eighteen months after the period they describe. Bitcoin is continuous. UBWI must therefore hold three timestamps apart and publish all three:

the Bitcoin observation timestamp, the instant the numerator was observed;
the denominator vintage, identified by its reference date, its publishing source, and its publication date;
the index calculation timestamp, the instant UBWI was computed.

Between denominator updates the denominator is held fixed at the latest published vintage. Urdais does not interpolate a denominator, nowcast it, extrapolate it by an asset-price proxy, or roll it forward by a growth assumption. UBWI between vintages moves only with Bitcoin. This is the correct behaviour for a share of a stock that is genuinely measured once a year: a denominator that moved daily would be a model, and the product would then be reporting the model rather than the measurement.

A new vintage takes effect from its own publication date forward. Values already published are not restated when a vintage is superseded, because they were correct statements of the best measurement available when they were made. Errors are corrected under the revision policy; superseded vintages are not errors.

Percentage change is withheld across a vintage boundary. A vintage change moves the denominator by several percent in one step, and a change computed across that step would describe the arrival of a statistical publication rather than any movement in Bitcoin or in wealth. Change is therefore computed only between values sharing the same denominator vintage and the same methodology version, and is withheld across either boundary rather than shown as a jump or as zero.

Major-Economy Treatment

The requirement is disclosure, not observation. An earlier proposal required every economy above 3 % of world GDP to be observed or the index would not publish. That is a permanent block on China, which does not compile the statistic, and a rule that can never be satisfied is a decision never to publish taken silently rather than a standard.

Instead: every economy above the disclosure threshold that UBWI does not observe is named on the published surface, with its GDP weight and the reason it is unobserved, and the published sensitivity range spans a plausible range of its wealth-to-GDP ratio. Phase 2C measured the China band at 0.2440 %–0.2927 % — roughly 17 % of the value from one economy's assumed ratio — which is a far more informative disclosure than any coverage percentage.

Sensitivity Methodology

The published range is not a confidence interval. No distribution over world wealth is available and none is assumed. The range is the span of four named, reproducible assumptions about the single parameter that dominates the uncertainty — the wealth-to-GDP ratio of the world UBWI cannot observe:

Unobserved world as wealthy per unit GDP as the observed set: tail ratio $R$.
Central, CWON-2020 calibrated tail: tail ratio $k \cdot R$. This is the published value.
Low: tail ratio $0.60 \cdot R$.
High: tail ratio $1.15 \cdot R$.

Each scenario is stored with its ratio, its imputed subtotal, its Total Global Wealth and its resulting UBWI, so the range is reproducible from the versioned assumptions alone. The central scenario is the published value; the others are diagnostics and an alternative denominator never becomes a second UBWI.

The range is published with the value, never in an appendix. Numerator uncertainty is not in it, because it does not belong there: the reference feed's 0.5 % deviation threshold bounds the price leg's departure from the wider market at 50 basis points, against a denominator sensitivity range spanning roughly a quarter of the published value.

Historical Reconstruction

UBWI may be reconstructed annually, at year-end, from 31 December 2013.

The binding constraint is the numerator, not the denominator. Bitcoin supply is reconstructible from the chain to the genesis block, and world-wealth vintages exist annually from the mid-1990s. But the Chainlink BTC/USD feed on Ethereum mainnet does not extend backwards indefinitely, and before it existed no price can be formed under this version's rule at all. A price that cannot be formed under the methodology's own rule is not a price the methodology may use.

This is a real narrowing relative to version 1.0.0, which could in principle have reconstructed a venue median back to 2013, and it is stated rather than glossed: the price amendment bought a publishable rights posture at the cost of reconstructible history. Reconstruction before the feed's first round would require a separate, versioned historical price rule, which does not exist. Points computable from a single-venue price are not published, because doing so would present a different construction under the same name.

Nothing is backfilled. No point exists for a year in which either the numerator rule or a denominator vintage cannot be satisfied. A gap is shown as a gap.

History is assembled vintage by vintage. Each historical point uses the world-wealth vintage in force for its own reference date, not the latest vintage applied backwards. Applying a current denominator to a past numerator would produce a series that never existed.

Source Hierarchy

In descending order of preference for the denominator:

A national-accounts compiler's own published world or multi-economy balance-sheet aggregate, at market value, in United States dollars at market exchange rates. No such publication is known to exist.
An open, documented, redistributable multi-economy database implementing the SNA national-wealth identity, from which a world aggregate is constructed under a stated rule. This is the V1 basis.
An official statistical agency, central bank, or international organisation dataset covering a defined subset of economies, used for cross-checking and for composition.
A published research estimate from a credible institution, used as a cross-check only, never as the denominator, and never where its terms forbid the use.
A commercial market-research estimate of a single asset class, used only as a non-additive cross-check and only with attribution.

A source enters the denominator only with a permitted and reproducible collection path and permitted data use, recorded as a permission basis, exactly as for every other Urdais output. Where a source requires attribution, the attribution is carried on every published surface.

For the numerator the hierarchy is shorter: a public onchain reference feed read directly by Urdais through its documented interface is preferred to any vendor aggregate, and both are preferred to a source whose terms forbid the use.

Chainlink rights posture: inferred permission, not a grant

Urdais's right to use the Chainlink BTC/USD Data Feed as a UBWI input is recorded as inferred_permitted. It is not expressly_granted, not licensed, and not confirmed_by_chainlink. The distinction is carried in the rights model itself, not only in prose: an inferred permission may support a numerator and may never supply a denominator constituent, and it can never be promoted to cleared by any code path.

What the inference rests on. The feed is exposed through Chainlink's documented public interface; the data is published onchain by the oracle network and readable by any Ethereum node, and Urdais reads it through public RPC endpoints requiring no key, no account and no click-through agreement; Urdais is not reselling or redistributing the raw feed; and Urdais uses one observed reference price as an input to its own derived wealth index.

What was not found, in both directions. No explicit prohibition of the Coinbase or Kraken kind — on caching, on automated recording, or on creating an external financial index from the data — was found in any Chainlink document Urdais retrieved. No explicit affirmative derived-index grant was found either. And Chainlink's terms could not be read at all: https://chain.link/terms, the document Chainlink's own documentation footer names as governing, returns HTTP 200 with a client-rendered shell carrying 3,441 characters of navigation text and no terms prose. Nothing in this methodology is quoted from a document Urdais has not read.

What the inference does not cover. It is not a licence; Chainlink has granted Urdais nothing, has not been contacted, and no outreach was sent. It extends to reading one reference price as an index input, and not to redistributing or mirroring the feed, nor to presenting Urdais's output as a Chainlink product. The terms of the upstream data providers whose data the oracle network aggregates are undisclosed to Urdais and are not covered. One successful retrieval of Chainlink's terms text settles the question either way and supersedes the inference.

The published surface carries this qualification with the price provenance rather than implying a licence.

Published Surface

One value. There are no UBWI variants.

Headline: the UBWI percentage; the Bitcoin market capitalization in United States dollars; Total Global Wealth in United States dollars; the denominator vintage with its reference date, source and publication date; the Bitcoin observation timestamp; the index calculation timestamp; the status; and the methodology version.

Beneath the headline, the denominator components, so that a reader can see what "global wealth" means here. Two panels, and they are never merged:

Additive components, which sum to Total Global Wealth: the sector decomposition of world net worth — net personal wealth, net non-profit wealth, net public wealth — plus the Bitcoin adjustment, plus the unexplained residual, shown rather than absorbed.
Non-additive cross-checks, each with its own source, reference date and valuation basis, and each labelled as overlapping: world real-estate market value, produced capital, natural capital, above-ground gold, and any single-asset-class aggregate. This panel carries a standing statement that its rows overlap and must not be added.

Percentage change where it is defined, and the reason it is withheld where it is not.

UBWI is displayed as a percentage with four decimal places. It is never displayed as a level, never given a base date, and never charted against index-point series without its unit stated.

Publication Gates

No UBWI value is published unless all of the following hold. The thresholds are configuration checked against the measured feasible frontier, not literals chosen for how they sound.

Every directly observed constituent is rights-cleared. Required, no exceptions.
Modelled share of the wealth denominator: at most 40 %.
Rights-cleared observed GDP coverage: at least 52 %.
No silently interpolated national wealth stock. Required.
Vintage age at the calculation date: at most 4 years.
Vintage dispersion, oldest to newest: at most 4 years.
Consumer durables included but not stripped: prohibited.
Sensitivity output available. Required.
Methodology version explicit. Required.
Denominator model version explicit. Required.
Source lineage complete, per component. Required.
FX lineage complete and end-period, per component. Required.
Every unobserved economy above 3 % of world GDP disclosed. Required.
Every numerator source rights-cleared or explicitly inferred-permitted, for the use actually performed — retaining the reading, deriving a commercial index from it, and displaying it. Required, and it covers the supply leg as well as the price leg. under_review and blocked both refuse.
Numerator observation valid, including the price feed's network, identity, round and staleness checks. Required.
Terms-artifact integrity. Every rights state must cite a retained document whose hash and byte length match the actual bytes and every word of whose quoted clause occurs, in order, in the document. An inferred permission must additionally state its basis and its limits. Required.

Why 40 % and 52 %, and why they are the same bound. A gate must be set against the feasible frontier. A threshold above it is not a high standard; it is a permanent refusal disguised as one. Phases 2A and 2B both proposed a 25 % imputed-share ceiling, which requires 68.44 % observed coverage; a 35 % ceiling requires 57.31 %. Both exceed the measured 55.72 % frontier, so either would have made the index refuse every denominator the global statistical system can produce, forever, while appearing to encode a quality rule. ≤ 40 % is the tightest satisfiable bound. The 52 % coverage floor is the same constraint stated in the other unit and moves with it; it is derived, not chosen. A configured floor above the stored frontier is itself refused as a configuration error.

An inferred permission satisfies the numerator rights condition and nothing else. It may never supply a denominator constituent, and it can never be promoted to cleared. This is the only widening of any gate condition in version 1.1.0, it is scoped to one explicitly decided source, and no threshold was changed.

The gate is never relaxed to make a calculation pass. Where any gate fails, no value is published and no substitute is shown. The failure is reported with the measured figure against the threshold.

As at version 1.2.0's effective date the gate measures a modelled share of 39.76 % against the 40 % ceiling and rights-cleared observed GDP coverage of 52.95 % against the 52 % floor, and the numerator's last external rights condition is gone — not cleared, but removed, the supply quantity now being derived from the protocol schedule. The gate passes with no finding, and the first UBWI value is published under this version. No threshold was changed to reach it.

Two gate conditions replace the retired supply-source rights requirement, and both are checks the gate performs itself rather than assertions about a third party:

the reference block height is valid and cross-verified — two independent observations, both non-negative integers, agreeing exactly, failing closed on disagreement; and
the protocol supply derivation is valid — the published supply is recomputed from the reference height and must reproduce exactly.

These are reported under four distinct findings — BLOCK_HEIGHT_INVALID, BLOCK_HEIGHT_NOT_CROSS_VERIFIED, SUPPLY_DERIVATION_INVALID and SUPPLY_IMPOSSIBLE — because an operator's response to two sources disagreeing has nothing in common with a supply above the protocol cap.

Update Cadence

The numerator is re-observed on each production calculation; it is instantaneous and its timestamp is load-bearing. The denominator is re-constructed when a constituent compiler publishes a new balance sheet, which is annual for every economy in the observed set except Canada, which publishes quarterly and is selected by a stored rule: the latest published observation at or before 31 December of the latest complete calendar year.

Between denominator vintages the denominator is held fixed and UBWI moves only with Bitcoin. Urdais does not nowcast, interpolate or roll forward a denominator. Percentage change is withheld across a vintage or methodology boundary, and is unavailable entirely until a second real observation exists: there is no such thing as a change from nothing, and no historical UBWI is fabricated to create one.

Revisions and Corrections

A superseded vintage is not a correction. When a new world-wealth vintage supersedes an earlier one, previously published UBWI values stand; the new vintage governs from its publication date.

A correction is a new interpretation, never an edit. Where an input is found to have been wrong — a misread supply figure, a venue price later repudiated, a denominator component mis-stripped — the affected values are superseded by new values carrying the correction reason, and the original values remain retrievable. Urdais does not silently restate a published number.

A methodology change never rewrites history. Values dated before a version's effective date remain those computed under the version in force at the time. Percentage change is withheld across the boundary.

Known Limitations

Stated with every published value.

The denominator is materially narrower than "all wealth". It excludes human capital by design, and it excludes, by measurement failure rather than by design: crypto assets other than Bitcoin; collectibles and art outside what national compilers capture as valuables; informally held and undocumented assets, which are a larger share of wealth in lower-income economies; and the going-concern premium that markets place on businesses beyond the assets on their balance sheets, which the SNA recognises as an asset only when evidenced by a sale.

The denominator is stale by construction. Its reference date is a year-end one to two years before the Bitcoin observation. UBWI is therefore a ratio between a current numerator and a lagged denominator, and it is labelled as one.

Coverage is uneven and partly modelled, and the world's measured net foreign position does not net to zero.

The exchange-rate conversion has a known direction of error. Where a purchasing-power-denominated wealth aggregate is converted to market exchange rates using an economy-weighted factor, the conversion over-deflates, because wealth is more concentrated in high-income economies than income is. The resulting denominator is a lower bound and the resulting UBWI an upper bound.

Uncertainty in the denominator is larger than any plausible uncertainty in the numerator. The numerator's price leg is bounded by the feed's 0.5 % deviation threshold and its 3600-second heartbeat. Reasonable denominator choices move UBWI by a factor approaching two. Any presentation that implies four-decimal precision in the underlying quantity, rather than in the arithmetic, is misleading, and the published sensitivity range exists to prevent that reading. The honest form of the answer is a range, not a point.

Open Questions

Resolved since 0.1.0-draft: the denominator source-rights blocker (every constituent is now cleared against a retained, hashed terms artifact); which world aggregate to use (none exists, so UBWI constructs one); and the exchange-rate conversion (end-period reference rates per component, replacing the purchasing-power conversion and its known bias).

Open, and each is recorded rather than worked around.

~~The Bitcoin supply source is not cleared for the use UBWI makes of it.~~ Closed in version 1.2.0, and closed by removal rather than by clearance. Blockchain.com's terms are unchanged and still do not grant the derived-index use; version 1.2.0 stops depending on them, deriving the supply from the block subsidy schedule instead of retrieving it. Version 1.1.0's note that issued supply is "a deterministic property of the chain … reproducible from the issuance schedule at a stated height" was correct, and 1.2.0 acted on it. What replaces the rights condition is stricter than it: the gate now recomputes the supply from the reference height and refuses if it does not reproduce.

What this closure does not claim: it is not a finding that Blockchain.com permits the use, and the construction did change — from claimed issuance to scheduled issuance, a difference estimated at roughly 44 BTC and disclosed under "Supply" as an estimate inherited from the retired source.
The imputed-share ceiling is met, but narrowly. Taiwan's admission brings the modelled share to 39.76 % against the 40 % ceiling, a margin of 0.24 percentage points. A revision to any large component, or a world-GDP restatement, could move it back above. Further rights-cleared coverage is still the durable answer.
Six near-frontier economies — Norway, Finland, Hungary, Israel, Latvia, Portugal — are worth 1.73 pp of coverage between the achieved figure and the 55.72 % frontier. Finland entered this version; the others are each blocked on a matched net-foreign-position year or, for Norway, on a non-produced asset series that stops in 2014. Admitting one without its land valuation would mean treating fixed capital stock as national net wealth, which this methodology prohibits.
The composition-sensitivity band is approximately 16 pp against a proposed ≤ 15 pp target, reachable at roughly 58–60 % coverage.
Chainlink's own terms could not be read. The rights state for the price leg is an inference from a documented public interface and from the absence of a retrieved prohibition, not a grant. One successful retrieval of the terms text settles it either way. Upstream data-provider terms remain undisclosed and are outside the inference.
An operational Bank of Korea ECOS API key for production retrieval volumes. Korea's publication rights are cleared and its value is manually verified against the first-party table; only the automated collection path is pending. These are separate questions and only the second is open.
Non-Bitcoin crypto assets in the denominator, required by the asset boundary and still unmeasured under a redistributable licence.
A second independent price reference. Version 1.1.0 depends on one feed. A documented eligibility rule for a second onchain reference, and a rule for reconciling the two without moving the series, is not yet written; adding one is a versioned change.

Relationship to the Urdais Product Surface

UBWI appears in the Urdais market catalog as an index symbol. Its unit is a percentage of Total Global Wealth. Any existing representation of UBWI as a points series with a base value is placeholder data and does not reflect this methodology.

Methodology Version

1.2.0, 15 September 2026. Status: approved for production, effective 15 September 2026. The residual denominator model carries its own version, 1.0.0, unchanged by this amendment, and a published value is immutable under both: a correction is a new, superseding publication, never an edit.

Version History

1.2.0, 15 September 2026: numerator supply amendment. Replaces externally reported claimed issuance, read from Blockchain.com's Explorer API, with protocol-scheduled cumulative block subsidy derived from the reference block height — 50 BTC initial subsidy, halving every 210,000 blocks by integer division, accumulated in integer satoshis, inclusive of the reference height, excluding transaction fees, with no lost-coin and no spendability adjustment. Records the supply rights basis as derived_from_protocol: no third-party data licence is relied upon for the quantity, so no claim about any provider's permission is made in either direction. Retires the external supply source rather than clearing it — Blockchain.com's terms are unchanged, still do not grant the derived-index use, and the retained artifact is preserved as the evidence for the retirement. Keeps Phase 2E's two-source block-height cross-check unchanged, with exact agreement and fail-closed on disagreement, and freezes each source's identity, raw value, retrieval timestamp and provenance on every published point. Replaces the gate's external-supply rights requirement with two stricter checks the gate performs itself — the reference height must be valid and cross-verified, and the supply must reproduce from that height by recomputation — under four distinct findings (BLOCK_HEIGHT_INVALID, BLOCK_HEIGHT_NOT_CROSS_VERIFIED, SUPPLY_DERIVATION_INVALID, SUPPLY_IMPOSSIBLE). No gate threshold was changed, and no gate was weakened: a rights requirement was removed because the dependency it governed was removed. The historical underclaimed-coinbase difference is disclosed and explicitly labelled an estimate carried over from the retired source. This is the first version under which a UBWI value is published.

1.1.0, 15 September 2026: numerator price amendment and one denominator admission. Replaces the three-venue Coinbase/Bitstamp/Kraken spot median with the Ethereum mainnet Chainlink BTC/USD Data Feed reference price, pinned to proxy 0xF4030086522a5bEEa4988F8cA5B36dbC97BeE88c, with a fail-closed validator over network, feed identity, round and a 3600-second heartbeat staleness bound, and with full proxy, phase, aggregator, round and block lineage frozen on every published point. Distinguishes the proxy as the methodology pin from the aggregator as observation lineage, so an ordinary aggregator upgrade is not a methodology break while a change of proxy, pair, network or product type fails closed. Records the Chainlink rights state as inferred permission, not an express grant, with its basis, what was searched for and not found in both directions, and what the inference does not cover; introduces inferred_permitted as a source-rights state that may support a numerator and may never supply a denominator constituent. Admits Taiwan to the observed set on DGBAS National Wealth Statistics for 2024 under the Open Government Data License, Taiwan 1.0, with consumer durables stripped, land included at announced current land value, and the World Bank's own statement on Taiwan's treatment in WDI aggregates as the basis for the residual handling. The denominator now satisfies the 40 % modelled-share ceiling (39.76 %) and the 52 % coverage floor (52.95 %); no threshold was changed. The retired venue observation, and every retained venue terms artifact behind its retirement, are preserved rather than deleted. No value is published under this version at its effective date: the Bitcoin supply source's terms do not grant the derived-index publication Urdais performs, that source is a numerator source, and the gate refuses.

1.0.0, 15 September 2026: first production methodology. Defines Total Global Wealth as a constructed hybrid of directly observed, rights-cleared national balance sheets and a versioned modelled residual for economies that do not compile a comparable balance sheet, and states why that modelling is structurally necessary rather than provisional. Adds the production denominator construction, the residual model and its CWON-2020 calibration, the four-scenario sensitivity methodology, the end-period per-component FX policy replacing the purchasing-power conversion, the major-economy disclosure rule replacing the unsatisfiable major-economy observation rule, the concrete publication gate with thresholds checked against a measured feasible frontier, the vintage age and dispersion bounds, and the update cadence. Records that the Bank of Korea component is manually verified and its automated retrieval pending, and that numerator source terms are unreviewed. Applies the vintage rule, which removes New Zealand (2017) and Russia (2019) from the observed set rather than bridging them. No value is published under this version at its effective date: the modelled share is above the 40 % ceiling and the gate refuses.

0.1.0-draft, 14 September 2026: initial methodology. Fixed the primary question and the percentage unit; defined the numerator as claimed issued supply times a three-venue median spot price with no lost-coin adjustment; defined Total Global Wealth as consolidated world net worth on the SNA national-wealth identity, excluding human capital and consumer durables, including government assets and liabilities, natural capital, valuables and monetary gold, and including crypto assets without a corresponding liability; stated the anti-double-counting rule and its four prohibitions; selected national balance sheets over household net worth, comprehensive wealth and asset-class aggregation; placed Bitcoin inside its own denominator; fixed the vintage policy, the three timestamps and the withholding of change across a vintage boundary; set historical reconstruction at 31 December 2013 on a numerator constraint; and recorded the source-rights blocker that then prevented publication.