Image credit: Unsplash
Estimates of how much value moves across Bitcoin can differ by as much as six times depending on how transaction outputs are counted, a Bank for International Settlements working paper published Sept. 15 found.
The difference largely comes from deciding which outputs represent economic transfers and which simply return Bitcoin to the sender as change. The study examined Bitcoin, Ethereum and Tron using data from Mercurius.
Bitcoin Change Outputs Create Sixfold Transfer-Value Gap
Bitcoin uses unspent transaction outputs rather than account balances. When a user spends only part of an existing output, the remainder is typically returned as change, but raw blockchain data does not identify that output automatically.
Researchers compared three approaches to calculating Bitcoin transfer value. The highest estimate counts all transaction outputs.
A second method removes outputs identified as returning to sending addresses. A lower estimate also excludes the largest output when no self-transfer can be identified, treating it as likely change.
The resulting estimates differed by as much as sixfold. The Bitcoin dataset covers activity from 2009 through 2026 and includes about 1.3 billion transactions and 3.6 billion outputs.
Bitcoin Market-Cap Methods Produce Fourfold Valuation Differences
The paper found similar methodological differences in measures beyond transaction value. Bitcoin’s conventional market capitalization has at times reached roughly four times its realized capitalization.
Realized capitalization instead values Bitcoin’s UTXO supply using the market price associated with when each output last moved. The paper does not describe either measure as inherently wrong. Instead, they answer different questions and can produce substantially different pictures of network value.
Ethereum and Tron Reinforce BIS Methodology Warning
The researchers found that measurement choices also affect how activity is interpreted across Ethereum and Tron.
Mercurius contains roughly 100 billion stored records across the three networks, but the paper cautions that those records should not be treated as 100 billion separate blockchain transactions because the same information can appear at multiple processing stages.
The BIS paper does not argue that public blockchain data is unreliable. Its central conclusion is that on-chain indicators remain estimates shaped by definitions and methodology rather than direct measures of economic activity.
