Bitcoin onchain transfers vary sixfold in BIS study

A Bank for International Settlements working paper has found that estimates of Bitcoin onchain transfer value can differ by as much as sixfold depending on how researchers treat transaction outputs.

The Bank for International Settlements working paper, titled Hidden by complexity? Measuring stablecoin, crypto and decentralised finance ecosystems, was written by Timothy Aerts, Ronald Heijmans, Jan Paulick and Violeta Vuletic and published in September 2026. The research uses data from Bitcoin, Ethereum and Tron collected through Mercurius, a project operated by De Nederlandsche Bank and developed with the BIS Innovation Hub and Deutsche Bundesbank. Read BIS Working Paper 1377

Across the three networks, the Mercurius dataset contains 100 billion records. The paper cautions that the figure represents data points stored through its processing system and should not be read as 100 billion distinct blockchain events. Ethereum and Tron records can overlap between base-layer and smart-contract data, while information can appear at several stages of the processing pipeline.

Bitcoin onchain transfer estimates depend on change handling

For its Bitcoin analysis, the study examined data covering 2009 through 2026, including approximately 1.3 billion transactions and 3.6 billion transaction outputs. Researchers tested three ways of calculating transfer value because Bitcoin’s unspent transaction output, or UTXO, model can return unused funds to the sender as change.

The paper’s upper estimate counts the value of all outputs. An adjusted measure removes outputs sent back to the sending address, which researchers treated as likely change. A conservative lower estimate removes identified self-transfers or, when none can be identified, subtracts the transaction’s largest output under the assumption that smaller outputs represent the economic transfer.

Applying the different methods produced gaps of as much as sixfold in monthly estimates of Bitcoin onchain transfer value. Figure 5 in the study shows raw, adjusted and lower-bound estimates diverging sharply during several periods, especially when Bitcoin activity and prices increased.

Researchers cautioned that their lowest estimate is a conservative heuristic, not a mathematically certain measure. CoinJoin transactions can contain several change outputs, while mixers, spam activity and intermediary transfers create other classification problems that the model does not attempt to remove. The paper found that self-transfer exclusions became more pronounced from March 2016 as address reuse increased.

The authors said commonly quoted metrics can convey “a degree of accuracy that is not supported by the nature of the underlying data.”

Bitcoin market cap changes under alternate valuation methods

The same measurement problem appears when valuing Bitcoin’s supply. The study compared conventional market capitalization with measures that account for long-dormant coins and the price at which individual outputs last moved.

Standard market capitalization applies the current Bitcoin price to all outstanding supply. One alternative in the paper removes UTXOs that have remained untouched for more than 15 years. Using that threshold excluded just over 1.8 million BTC from the calculation, although the researchers stressed that inactivity cannot prove private keys have been lost.

Their data show approximately 3.5 million BTC had remained dormant for more than 10 years, while 1.8 million BTC had not moved for more than 15 years. Yet nearly 24,000 BTC eventually moved after sitting dormant for over a decade, including close to 3,000 BTC that moved after 15 years. The researchers cited those movements as evidence that age-based lost-coin estimates remain uncertain.

A separate measure, realized capitalization, values each unspent output using the Bitcoin price recorded when that output was created after its previous spend. The BIS analysis found conventional market capitalization reached as much as four times realized capitalization during periods of rapid price appreciation. During Bitcoin’s sharp 2022 decline, realized capitalization temporarily stood above conventional market capitalization because older outputs retained their previous valuation.

In related coverage, a crypto.news explainer on Bitcoin realized price describes realized capitalization as the basis for calculating the market’s aggregate onchain cost basis. Crypto.news reported in August that Bitcoin’s realized capitalization rose by more than $4.6 billion during one seven-day period, based on CryptoQuant data. Read the related Bitcoin realized-cap report

Ethereum and USDT create different data problems

Ethereum presented a different challenge because a single transaction can call several smart contracts and generate information across transaction inputs, execution traces and event logs. The BIS researchers examined 67.5 million deployed and active Ethereum contracts and could not place more than 54 million into their technical categories.

Among contracts they could classify, close to 12 million were proxies, 1.4 million were fungible-token contracts and approximately 100,000 were NFT contracts. Researchers used bytecode and standards such as ERC-20 and ERC-721 for classification but warned that adherence to a technical standard does not establish a contract’s economic purpose.

Naming created another source of noise. The paper identified the USDT symbol across roughly 7,000 Ethereum token contracts even though only one represented Tether’s official Ethereum issuance contract. Its analysis excluded the genuine Tether contract when studying imitation or potentially spurious USDT-labelled tokens.

Cross-chain USDT data showed a separate contrast. On Ethereum, the proportion of USDT held in smart contracts climbed above 20% in 2022 and stayed mostly between 15% and 20% until late 2024 before falling into a roughly 10%–15% range. On Tron, smart-contract accounts held around 1% of USDT during most periods examined.

The researchers associated Ethereum’s smart-contract balances with activities including liquidity provision, lending and other DeFi uses. They described Tron USDT as appearing more frequently in transactional and store-of-value activity, while warning that account type remains an imperfect proxy because externally owned addresses can represent payments, exchange custody, remittances or holdings.

In related coverage, crypto.news reported that Tron processed $2.1 trillion in USDT transfers during the second quarter of 2026, citing Messari data, while its USDT supply ended the quarter at $87.9 billion. Read the crypto.news Tron USDT report

BIS favors ranges as Visa filters stablecoin activity

For future measurement work, the authors recommend supplementing single-number estimates with bounded ranges that disclose protocol-specific uncertainty. Their framework calls for technical contract classification, explicit assumptions and separate treatment of an asset’s identity and the blockchain infrastructure on which it operates. The authors described onchain indicators as “noisy approximations rather than direct measures of economic activity.”

The publication is a BIS Working Paper and does not itself introduce a regulation, compliance deadline or reporting requirement. BIS states that Working Paper views belong to their authors and do not necessarily represent the institution or its member central banks.

A comparable filtering process is already visible in the Visa Onchain Analytics dashboard, developed with Allium Labs. Visa separates total stablecoin activity from an adjusted measure designed to filter high-frequency trading, bots, bridge routing, centralized exchange activity, minting and burning, and other transfers that may not resemble ordinary settlement.

A Sept. 16 dashboard snapshot carried by TradingView from Cointelegraph showed $6.4 trillion in total stablecoin transaction volume over the preceding 30 days and $313.1 billion after Visa’s adjustments. The live Visa methodology states that its single-direction filter counts only the largest stablecoin transfer within one transaction, reducing duplicate internal movements generated by complex smart-contract calls.

Visa’s address filter includes transfers in adjusted volume when the activity meets its economic criteria and an address has not sent more than 1,000 transactions or $10 million in volume during a 30-day period. The dashboard classifies activity across payments, DeFi, centralized exchanges, investment and trading, store of value, minting and burning, short-term routing and infrastructure, while transactions below $250 that pass its adjusted filter are classified as retail-sized.