BIS Working Paper Admits Its Own Crypto Metrics Are Noisy Approximations

The Bank for International Settlements published working paper 1377, titled “Hidden by complexity? Measuring stablecoin, crypto and decentralised finance ecosystems,” on 15 September 2026. According to CryptoTimes, the paper examined roughly 1.3 billion Bitcoin transactions and 3.6 billion transaction outputs recorded between 2009 and 2026, alongside Ethereum and Tron data, drawing on a dataset the researchers describe as covering about 100 billion blockchain records in total.

The central finding, reported by both BitKE and Whale Factor, is that estimates of Bitcoin on-chain transfer value can differ by as much as a factor of six depending purely on which methodology is chosen. The broadest measure simply sums every transaction output. A more adjusted measure strips out change returned to the sender’s own address. A conservative measure goes further and tries to remove additional self-transfers that do not represent a genuine economic exchange between two parties. Each choice produces a materially different monthly figure, according to ForkLog, which noted that the analysis concerns on-chain transfer volume rather than exchange trading volume.

Smart contracts and stablecoins compound the problem

The paper also examined Ethereum smart contract activity and found that a very large share of active contracts could not be reliably classified using the study’s own taxonomy, according to BitKE. Stablecoin measurement presented a further complication because the same token can serve different economic functions depending on which blockchain it sits on. The researchers found that USDT on Ethereum is more closely tied to decentralised finance activity conducted through smart contracts, while USDT on Tron behaves more like a payment or store-of-value instrument held directly in wallets, a distinction that Whale Factor and BitKE both highlighted as evidence that a single aggregated stablecoin figure can mask two very different kinds of usage.

The BIS researchers’ own conclusion, quoted by both CryptoTimes and ForkLog, is that “on-chain indicators should be treated as noisy approximations of economic activity, rather than precise measures.” ForkLog reported that the authors recommended analysts stop citing single point figures without disclosing the methodology behind them, and instead publish ranges bounded by explicit assumptions, segmented by blockchain architecture.

Why the admission matters for regulators

The BIS is not a peripheral commentator on crypto markets. It is the institution whose research arm has repeatedly informed the policy conversation around stablecoin oversight, tokenisation and the case for central bank digital currencies, often citing on-chain volume and market capitalisation figures as evidence of scale or risk. This paper’s admission that those same categories of figures can swing by a factor of four to six depending on the counting method raises an obvious question. If the BIS itself now says these numbers were never precise measures to begin with, how much of the earlier policy narrative built on comparing crypto’s raw transaction volume against traditional payment systems rested on a foundation the institution is only now flagging as shaky. The paper does not retract any specific past claim, and it does not name any programmable money initiative directly. But the timing, arriving in the middle of ongoing digital euro and CBDC design work at the BIS Innovation Hub, invites scrutiny of how selectively noisy metrics may have been used when they supported a particular argument and quietly caveated when they did not.