Counting the Crime on the Blockchain: How Much Bitcoin Is Illegal?

"Bitcoin is mostly used for crime" was one of the standard criticisms. "Bitcoin is barely used for crime, cash is far worse" was the standard rebuttal. Both sides asserted with total confidence, and neither had a number.

Sean Foley, Jonathan Karlsen and Tālis Putniņš produced one. The method they used is a small masterpiece of empirical creativity, and it is worth understanding even if you never think about crypto again, because it is a template for how to measure something that is deliberately hidden.

The problem: how do you measure something designed to be invisible

Criminals do not fill in surveys. There is no register of illegal transactions. Every estimate of the size of any black market, drugs, trafficking, money laundering, is a guess built on a guess.

Bitcoin, though, has a peculiar property. Every transaction ever made is permanently public. The blockchain is a complete, immutable record of every movement of every coin since 2009. What it does not contain is names. Addresses are pseudonymous strings, not identities.

So the data are all there, and the labels are all missing. The problem is not observation. It is identification.

The key idea via analogy: start with the known criminals and follow their friends

Here is the trick, and it borrows from how a detective would actually work.

You do not know who the criminals are. But you know a few. Specifically:

• Law enforcement seizures. When police shut down a darknet marketplace or arrest a dealer, they seize Bitcoin, and the addresses involved become known. Those addresses are confirmed criminal.
• Darknet marketplaces. The major illegal marketplaces (the Silk Road and its many successors) published Bitcoin addresses to receive payments. Those addresses are confirmed criminal.

So you now have a seed set: a collection of addresses you are confident belong to illegal activity.

Then you use the blockchain's public nature to work outward. If an address sent money to a known darknet market address, that address probably belongs to someone buying drugs. If an address received money from a known market's payout address, it probably belongs to someone selling. You trace transactions backwards and forwards from your seeds, and you use network analysis to cluster addresses that are likely controlled by the same person, along with a set of statistical methods to identify users whose transaction patterns resemble known illegal users even when there is no direct link.

In effect, you use the small set of people you can positively identify as criminals to learn the behavioural signature of a criminal user, and then you scan the entire network for that signature.

The numbers they produced were striking:

• Roughly one quarter of Bitcoin users were involved in illegal activity, by their estimates.
• Around forty-six percent of Bitcoin transactions were associated with illegal activity.
• The annual value involved was estimated at around seventy-six billion dollars, which the authors note is close in scale to the entire US and European market for illegal drugs.

Those are large enough numbers that the "crypto is for criminals" camp felt vindicated. But the paper's most interesting finding is the one that came next.

The illegal share was declining. As Bitcoin attracted mainstream interest, speculation and investment, the criminal use, while growing in absolute terms, shrank as a proportion of total activity. Ordinary speculators simply swamped the criminals in the data.

And the criminals were leaving. The illegal share also declined as more privacy-focused cryptocurrencies emerged. This makes perfect sense once you understand the method of the paper itself: Bitcoin is a terrible tool for crime precisely because the ledger is public. Everything you do is recorded forever, and as this paper demonstrates, a determined researcher with a few seed addresses can trace you. Criminals who understood this migrated to coins designed to hide the transaction graph.

There is a delicious irony in there. The paper that established Bitcoin's role in crime also explains why Bitcoin was never a good vehicle for it, and demonstrates that fact by doing the tracing.

Why it mattered

• It replaced assertion with measurement. The debate over criminal use of crypto had been entirely rhetorical. This paper made it empirical, in a top journal, with a transparent method.
• It is a landmark of blockchain forensics. The seed-and-trace approach, using confirmed criminal addresses to learn a signature and then scanning the network, is now the core method of the blockchain analytics industry. Firms whose entire business is tracing illicit crypto flows for law enforcement and exchanges use techniques in this family.
• It reframed the privacy debate. Bitcoin's transparency is a genuine liability for anyone doing something illegal, and a genuine asset for investigators. That is a much more useful frame than "crypto is anonymous," which was never true.
• It showed the character of the asset changing over time. Watching the criminal share fall as speculators arrived is a quantitative portrait of Bitcoin's transformation from a darknet payment rail into a speculative asset. It documents the transition with data rather than narrative.
• The method generalises. Any setting where you have a complete transaction record and a handful of labelled bad actors is amenable to this approach. That is a real methodological contribution well beyond crypto.

The honest limitations

• The estimates depend heavily on the clustering assumptions. How you decide that two addresses belong to the same user, and how you decide that a user "resembles" a known criminal, drives the headline numbers. Reasonable alternative choices produce meaningfully different figures, and other researchers using different methods have reported substantially lower estimates of illegal activity.
• Guilt by association is a real risk. An address that transacted with a darknet market might belong to a buyer, a researcher, an exchange, a mixer used by both criminals and privacy-conscious ordinary people, or someone who received coins that had passed through a market three hops earlier. The taint spreads through the graph in ways that do not perfectly map onto culpability.
• The definition of illegal is broad. The category sweeps together drug purchases, marketplace operators, and other unlawful activity, and reasonable people disagree about what belongs in that bucket.
• The estimates are for a specific era. The data cover Bitcoin's darknet-heavy period. Both the absolute and relative figures have moved substantially since, and applying these percentages to today's crypto market would be badly wrong. The paper's own finding of a declining trend tells you that.
• It does not compare against cash. The most common defence of crypto is that physical cash finances vastly more crime. That may well be true, and this paper does not address it, so the numbers should not be read as evidence that crypto is unusually criminal relative to alternatives.

The one-line takeaway

Foley, Karlsen and Putniņš turned Bitcoin's permanent public ledger against its criminal users, using seized addresses and darknet marketplaces as seeds to trace illegal activity through the transaction graph, and found that a large share of early Bitcoin activity was illegal, that the share was falling as speculators arrived, and that criminals were fleeing to more private coins precisely because Bitcoin is so easy to trace.