Tuesday, August 18, 2020

Bitcoin is an account, not a token

When economists talk about payments, they often make a distinction between token-based and account-based payment systems. In a recent post at the New York Fed's Liberty Street blog, Rodney Garratt & cowriters argue that new payments technologies like bitcoin and central bank digital currency may not fit into these traditional categories. Perhaps it's time for a reorg?

In an account-based system, some sort of database stores account information. For a payment to occur across this database the payer needs to prove that they own a spot in that database, and that this spot has sufficient funds.

With a token-based system there is no database. Instead, objects are used to pay (say banknotes or gold coins). The key feature of a token-based system is that the recipient must verify that the object is valid and not counterfeit.

In short, tokens involve identifying the object. Accounts involve identifying the individual.

Garratt et al argue that a digital currency such as bitcoin is a mix of the two, token and account. Bitcoin is an account-based system because some sort of "proof of identity," specifically a private key, is needed to transact. This puts bitcoin in the same category as a bank account, which also has a process for verifying the identity of users. (Instead of public key cryptography, bank customers must go through a due diligence process, and after that must produce a PIN.)

I think Garratt et al are right about bitcoin being an account-based system. But I don't think that bitcoin also qualifies as a token-based system.

Not a token

A token is an object. And objects can be counterfeited. That's why when Alice pays Bob a $20 note, Bob is responsible for carefully checking it.

Account-based systems don't suffer from counterfeiting problems. It's impossible for a Alice to pay Bob with fake dollars from her Wells Fargo account. Wells Fargo dollars aren't independent objects in the same way that banknotes are. They are entries in a well-secured database. The same goes for bitcoins. There is no way for Alice to make a fake bitcoin entry and send it to Bob.

In addition to counterfeitability, I'd argue that another key feature of a token is that it can be lost by one person, found by another, and then reused for payment. If Bob loses either his $20 banknote or his gold Krugerrand, Alice can find them and spend them.

But account-based systems don't have this feature. Wells Fargo dollars are database entries. It'd be impossible for Alice to lose a Wells Fargo database entry or Bob to find one and reuse it. And the same goes for bitcoins. Alice can't lose 0.2234 bitcoins, nor can Bob find Alice's 0.2234 bitcoins.

Gift cards: account or token?

What about gift cards, say like those iTunes cards that they sell at pharmacies?

Because gift cards come in physical form they may seem like tokens. But the card is just one element in an account-based system. A $20 iTunes balance exists in a database run by the gift card system operator, Apple. Whereas a $20 banknote can be used without the authorization of the central bank, an iTunes card has to be granted permission before it can initiate purchase. If the database entry to which an iTunes card is linked is empty, then a payment will be denied.

Like other account-based systems, counterfeiting isn't a problem with gift cards. Database entries can't be produced with an inkjet printer. (Sure, an iTunes card can be lost and reused by the finder. And so it seems like it should be classified as a token. But that's only because the "key word" or "password" is literally baked onto the card. Losing an iTunes card is like losing one's private bitcoin key.)

Bitcoins and gift cards can be turned into tokens

There is an interesting situation in which gift cards or bitcoins can be converted from account-based systems into token-based systems.

Instead of using her $20 banknote to pay Bob, Alice may choose to pay Bob by passing him a $20 iTunes card. Or maybe she gives him a physical Opendime hardware wallet that holds $20 in bitcoin. Bob could in turn pass these devices on to Charlie, to whom he owes $20, and Charlie on to Dave, etc.

In both cases, accounts are being repurposed as tokens. What is happening is the physical object, or key, that provides access to underlying bitcoin or gift card database entries is being moved from one person to another. But the actual database entries to which the Opendime and iTunes card are linked are not being updated.

Bob will have to treat both of these proffered instruments like cash. He'd be worried about counterfeits, and would want to verify that neither the gift card nor the Opendime device are fake.

Let's vizualize all of this into a table:

The table suggests that bitcoin exists in a totally different category than a banknote (unless those bitcoins are embodied in the form of an Opendime device). You'll also notice that the table differentiates between open accounts and closed ones. Let's get into that next...

Bitcoin and Wells Fargo are different types of accounts-based systems

If bitcoin qualifies as an account-based system, it is certainly different from Wells Fargo's system. I'd argue that Wells Fargo operates a "closed account" system while Bitcoin functions as an "open account" system. The key feature of an open system is that anyone can get an account. As Jerry Brito would put it, they are permissionless. A passport or driver's license isn't required to gain access, so even fraudsters can climb aboard.

Gift cards, non-reloadable prepaid cards (like Vanilla cards), and e-gold (a 1990-2000s era payments system that didn't require people to use real names) are also open account systems. Anyone can self-initiate account opening and start to make payments. No one gets kicked off. (We could further differentiate between centralized and decentralized open account systems, with gift cards being the former and bitcoin the latter.)

What qualifies something as a closed account system is that not just anyone can get access. Want to use a Wells Fargo account for making payments? You'll have to make it through Wells Fargo's application process, which involves giving up plenty of personal information. And only when a Wells Fargo employee has opened an account can payments be made.

Turning bitcoins, gold, and cash into closed account systems

Incidentally, it's also possible to convert tokens and open accounts into closed accounts. The London Bullion Market Association (LBMA) manages a walled garden of gold bars. Each bar is carefully vetted prior to entry into the system, and once in the system all movements are tracked by monitoring serial numbers. LMBA gold has ceased to be a token and has become a closed account. (I recently wrote about the LBMA system for Coindesk.)

Central bans could do the same with cash. They could set up an online bank note registry, and all cash users would be required to sign-up and record the serial numbers of note each time they receive a new one.

As for bitcoin, many people already keep their bitcoin in at regulated services like Coinbase that require customer ID. One day it may only be possible to send bitcoins from one regulated service to another, in which case a big chunk of the Bitcoin network will have become a closed account system, not an open one.

Why does all this matter?

Specialists need to be able to have conversations about their subject matter. Categorization is one of the ways to make these conversations flow without chaos. Are the categories we use for conversing on the topic of the economics of payments still sufficient? The world has changed. We've got new entrants like bitcoin and central bank digital currency. Garratt et al suggest that aging categories can "slow down progress in understanding intrinsic differences between the growing set of digital payment options and technologies."

But this case I think the old taxonomy are still useful, albeit with some tweaks.

P.S. Lawyers, regulators, users, and developers will all have different taxonomies for payments. Inter-taxonomic conversations are difficult, since a single term can mean something totally different. So stay within your taxonomy to avoid confusion.


  1. This is indeed an helpful framing. However, I think an essential feature of a token is the whole notion of being able to transact on a peer-to-peer basis. That was possible earlier with only notes, coins etc but only in person. Bitcoin etc make it possible to do the same digitally. Though at this point, there are all sort of intermediaries involved in a Bitcoin based transaction, but with adequate technical know how one can have a true peer-to-peer experience. I think the categorization framework does not fully take into account this key capability enabled by Bitcoin. By definition an means some form of liability or custodianship by the institution maintaining the account to the account holder. It is indeed right to see Coinbase an the likes as account maintaining / custodian institutions, but that is missing the point. A Bitcoin holder chooses to use their service, but he/she can with adequate technical know how manage it on his/her own. This is simply not possible in all the other examples discussed.

    1. Is bitcoin actually peer-to-peer? We can stand in front of each other and make some changes to the bitcoin database. But it's no different than standing in front of each other and using our accounts at Cash App or Zelle or Swish to make changes to banking databases.

    2. If I write my private key down on a piece of paper and hand it to you, then you can check the 'balance' using your bitcoin node and querying the blockchain (or via a block explorer).

      This transaction is therefore like any cash transaction and I think could accurately be described as peer-to-peer.

    3. "If I write my private key down on a piece of paper and hand it to you, then you can check the 'balance' using "

      Private key? Don't you mean public key? If I give someone my public key, they can check its balance.

      But I don't see why this qualifies something as peer to peer.

    4. The private key can be exchanged peer-to-peer just as cash is e.g you can hand this between two people in a dark alley, just as you would with a £20 note.

      As long as that private key has access to some unspent transaction outputs then is this not a peer-to-peer transaction just in the same way that cash is?

    5. Ok, I see what you mean. Yes, agreed. An Opendime device, which I wrote about in my article, is basically a bitcoin private key that can be traded from hand to hand like cash ie peer-to-peer. In my chart I put it in the token-based section.

      But if I send you some bitcoins to your address, that isn't peer-to-peer.

  2. Did cash not start out the same way, an account based system with the banks, (all cash that is created is accounted for), but with the pay bearer feature implicitly imbedding the password so that it could be (also) used as a token system?

    1. I'm not sure.

      Early Bank of England banknotes used to have people's names one them, and they were numbered.



  3. Thank you John Paul for the "stay within your taxonomy" warning! I fully share it with you. Too many lawyers quote economic concepts that ruin any legal approach to money.

    So I’ll try and be careful while escaping my legal approach by bringing up a few remarks.

    I wonder if the identity access to bitcoin is an economic issue: is it not more about technology (without being capable of saying what kind of technical knowledge it refers to)? Is it not the equivalent to numismatics for coins?

    Same goes with the argument about the re-use of bitcoins. I was convinced that if I found a USB key with a bitcoin wallet that was not protected, then I could use them. On the other hand, accounts can be cleaned when abandoned. Bitcoin stays in memory like a lost asset, a bit like a lost coin.
    What is interesting is also that the non-immediate re-use of bank money has also been a long time argument against legally recognising bank money as real money.

    I also find the counterfeiting argument does not lead to anything. It is an offense to a physical object. When it comes down to the core of the offense, counterfeiting is an act against the monetary sovereign (crime de lèse-majesté) and public trust.
    It has no purpose in a virtual world were copies can be perfect: “remember” when the point was to determine if a damaged coin was still a money? There is not such thing as a damaged bitcoin because a bitcoin is virtual asset.
    But if course, that does not mean that people can attack the bitcoin system in order to gain advantages and at the same time creating trust issues (market manipulation, forking).

    Finally, in my own taxonomy, I have not reached a conclusion if a bitcoin should belong to value monetary legislation (which could be token legislation) or account monetary legislation. I suppose that before that Bitcoin would have to make it from asset to currency.

    1. "I wonder if the identity access to bitcoin is an economic issue: is it not more about technology (without being capable of saying what kind of technical knowledge it refers to)? Is it not the equivalent to numismatics for coins?"

      I suppose it's not really a general economics issue. But within economics there is a narrow field sometimes referred to as the economics of payments systems. Folks like Kahn, Roberds, McAndrews, Garratt, and others are a few of the contributors. That's where the language I'm drawing from emerges, and they are interested in identity and anonymity.

      " I was convinced that if I found a USB key with a bitcoin wallet that was not protected, then I could use them. On the other hand, accounts can be cleaned when abandoned."

      In both cases you need to have the key or password to interact with the database. The account entries themselves can't get lost (nor can they be found later).

    2. On the blockchain there isn't actually a concept of a bitcoin, it's simply the reference to previously unspent transaction outputs.

      As such, if you found a private key which allowed you the ability to transfer previously unspent transaction outputs e.g had a 'balance' then you would be able to directly interact with the database.

      Drawing a parallel with cash, it's the difference between finding a wallet which has a bunch of bank notes in (e.g finding a private key) and finding a little transparent safe on the floor which you can see has money in but you can't open (e.g finding an address, an encrypted private key or looking on the blockchain).

      I therefore agree with Garratt et al whereby bitcoin is an account-based system but the private key itself feels like a token.

      Really keen to hear you thoughts on my response @jp Koning

    3. "As such, if you found a private key which allowed you the ability to transfer previously unspent transaction outputs e.g had a 'balance' then you would be able to directly interact with the database."

      Sure, but if I find someone's bank account information, I can directly interact with the database too. So are bank account passwords and login IDs also tokens? (It is a fact that login IDs and passwords get traded on underground markets).

      I'd agree that passwords and IDs can be tokens, but the underlying account itself is always an account-based product.

    4. But there's a difference in finding someone's sort code/account number vs their card with pin code written done. The former doesn't give me access to move funds whereas the later does.

      A private key gives you access to move any associated funds and therefore directly interact with the database to change ownership of something, just like finding cash on the ground or a card and pin number allows you to take ownership of those funds.

      It seems to be then that a token gives the user direct access and the ability to change ownership whereas an address/public key or logID only provides account information.