Tuesday, June 20, 2017

The road to sound digital money

No, I'm not talking about sound money in the sense of having a stable value. I'm talking about money that is sound because it can survive natural disasters, human error, terrorist attacks, and invasions.

Kermit Schoenholtz & Stephen Cecchetti, Tony Yates, and Michael Bordo & Andrew Levin (pdf) have all recently written about the idea of CBDC, or central bank digital currency, a new type of central bank-issued money for use by the public that may eventually displace banknotes and coin. Unlike private cryptocoins such as bitcoin, the value of CBDC would be fixed in nominal terms, so it would be very stable—much like a banknote.*

It's interesting to read how these macroeconomists envision the design of a potential CBDC. According to Schoenholtz & Cecchetti, central banks would provide "universal, unlimited access to deposit accounts." For Yates this means offering "existing digital account services to a wider group of entities." As for Levin and Bordo, they mention a similar format:
"Any individual, firm, or organization may hold funds electronically in a digital currency account at the central bank. This digital currency will be legal tender for all payment transactions, public and private. The central bank will process such payments by debiting the payer’s account and crediting the payee’s account; consequently, such payments can be practically instantaneous and costless as well as completely secure."
I don't want to pick on them too much, but all these authors are describing a particular implementation of central bank digital money: account-based digital money. There's an entirely different way to design a CBDC, as digital bearer tokens. My guess is that the authors omit this distinction because macroeconomists tend to abstract away from the differences between various types of money. Cash, coins, deposits, and cheques are all just a form of M in their equations. But if you get into the nitty gritty, bearer tokens and accounts two are very different beasts. Some thought needs to go into the relative merits and demerits of each implementation, especially if this new product is to replace banknotes at some hazy point in the future.

Let's first deal with account money. An owner of account-based money needs to establish a connection with the central issuer every time they want to make a payment. This connection allows vital information to flow, including instructions about how much money to transfer and to whom, confirmation that there is sufficient funds in the owner's account, and a password to confirm identity. Only then can the issuer dock the payor's account and credit the payee.

Bearer money, the best examples of which are banknotes and coins, never requires a connection between user and issuer. As I described in last week's post, courts have extended to banknotes the special status of having"currency." What this means is that if you are a shopkeeper, and someone uses stolen banknotes to buy something from you, even if the victim can prove the notes are stolen you do not have to give them back. The advantage of this is that there is never any need for a shopkeeper to call up the issuer in order to double check that the buyer is not a thief.** As for the issuer, say a central bank, they are not responsible for the debiting and crediting of banknote balances, effectively outsourcing this task to buyer and sellers who settle payments by moving banknotes from one person's hand to the other. The upshot of all this is that since users and issuers of bearer money don't need to exchange the sorts of information that are necessary for an account-based transaction to proceed, there is no need to ever link up.

This makes bearer money an incredibly robust form of money. If for any reason a connection can't be established between user and issuer, say because of a disaster or a malfunction, account-based money will be rendered useless. Examples of this include the recent two-day outage of Zimbabwe's account-based real-time gross settlement system due to excess usage, or the famous 2014 breakdown of the UK's CHAPS, its wholesale payments system, which limited the system to manual payments. M-Pesa, Kenya's mobile money service, has periodic outages, and last month my grocery store, Loblaw, suffered from a malfunction in its debit card system. Banknotes—which don't require constant communication with the mothership—worked fine throughout.

The private sector used to be heavily engaged in providing bearer money, both in the form of banknotes and bills of exchange. However, bills of exchange-as-money went extinct by the early 1900s. As for banknotes, the government thoroughly monopolized this activity by the mid-1900s. Which means the government has—perhaps inadvertently—taken on the mantle of being the sole issuer of stable, disaster-proof money. So any plan to slowly phase out government paper money is simultaneously a plan to phase out society's only truly robust payments option.

Going forward, it's always possible that governments once again allow the private sector to  issue bearer money. With the government's bearer money monopoly brought to an end, the public would be well-supplied with the stuff and central banks could safely exit the business of providing a robust payments option. But I can't see governments agreeing to relinquish this much control to private bankers. Which means that for society's sake, whatever digital replacement central banks choose to adopt in place of banknotes and coins should probably have bearer-like capabilities in order to replicate cash's robustness. Account-based money won't cut it. Nor will volatile private tokens like bitcoin.

One way to design a digital bearer money system is to have a central bank issue tokens onto a distributed ledger and peg their value, say like the Fedcoin idea. The task of verifying transactions and updating token balances would be outsourced to thousands of nodes located all over the world. So if all the nodes in the U.S. have been knocked out, there will still be nodes in Europe that can operate the payments system. This would restore a key feature of banknotes, that they have no central point of failure, thereby allowing central banks to get rid of cash. I'm sure there are other ways of creating robust money than using a distributed ledger, feel free to tell me about them in the comments section.

* CBDC would be redeemable on a 1:1 basis for traditional central bank money (and vice versa), so the two would have the same value and be interchangeable. Consumer prices, which are already expressed in terms of traditional central bank money, would now also be expressed in terms of CBDC. Since consumer prices tend to be sticky for around four months, CBDC holdings would have a long shelf-life. If CBDC was designed like bitcoin--i.e. its quantity was fixed and there was no peg to existing central bank money--then its value would diverge from traditional central bank money. Price would continue to be expressed in terms of traditional central bank money, and would be sticky, but there would be a distinct CBDC price that would no longer be sticky. So CBDC would no longer have a long-shelf life; indeed, CBDC prices could become quite volatile. See here.
** The caveat here is that while banknotes have long since been granted currency, CBDC—which does not exist—has not. Nor have cryptocurrencies like bitcoin been granted currency status. But if a central bank were to issue a bearer form of CBDC, it's hard to imagine the courts not declaring it to be currency fairly early on, unlike say bitcoin.

PS: I just stumbled on a 2006 paper from Charles Kahn and William Roberds which nicely captures these two types of money:


  1. Even digital bearer money is not quite as robust as paper and coin since it won't work without devices or power. Device loss may or may not be equivalent to paper and coin loss, but if not, it would still need cloud access and that may not be more robust than distributed central accounts.

    1. Yep, the need for power means that digital bearer money isn't as stable as good old paper money.

  2. All digital money is bearer cash because all cash devices can implement honest double entry accounting not requiring a third part ledger service.

    1. Can you explain that? I don't follow.

    2. Digital money requires the user have a local processor in hand. Since the local processor must be secure enough to hold a digital crypro coin, it thus has enough security to execute tamper proof honest accounting such that it never double spends. Put in other words.

      Right now I can double spend bitcoin, spending it a second time before the first goes through the ledger. But the new hardware wallets prevent that.

      For example, In China when the PBOC prohibited withdrawals for a time, traders simple passed the e codes for bitcoin between themselves with messaging software, which works fine as long as the two parties are trusted. But if the users have digital devices that have secure operating code,then any device can become trusted, just like a hundred dollar bill is generally trusted.

      Making trusted hardware wallets is now standard.

    3. "All digital money is bearer cash because all cash devices can implement honest double entry accounting not requiring a third part ledger service."

      What about mobile money, say something like M-Pesa?

    4. If you put the secure element inside the smart phone then M-Pesa can do direct wallet to wallet without using the M-Pesa central exchange. This van be done directly, card to card, person toperson using NFC, or it van be tunneled through the internet for direct wallet to wallet transfer. In this case, the actual M-Pesa exchange is used as a ledger service, only when needed,. Or the M-Pesa service can specify a limit on direct peer to peer exchanges before one call to he central service is needed.

    5. Interesting. But no one is actually doing this, are they?

    6. Since bitcoiners already pass ecodes over the email for direct person to person transfer, then I have a difficult time understanding how a secure bitcoin wallet would avoid offering the same capability. The user need only read out the bit coin code from the wallet, then send it along to another person who inserts it into their wallet. In other words, it would be an actual denial of the obvious.

    7. Also, I should mention that some hardware wallets are nodes on he block chain. Hence, there is no exchange service, the wallet submits the transaction, and the other wallet finds confirms the result. The intermediate block chain simply serving as a network. But, overall, the idea that two hardware wallets cannot pass crypto coins directly between them is absurd and defies all logic about what real cash and real wallets are all about.

  3. Here's my take:


    For context:



    1. Thanks for the links, Rohan, especially the second one which I hadn't seen before. Would you say that a central bank could deploy eCurrency Mint's product as either an account-based solution or a bearer solution, depending on their needs? Or is there a better way to think about eCurrency?

    2. I would say eCurrency Mint's product is definitely a bearer instrument first and foremost, but that you could set up an accounts-based payments intermediary layer on top that uses the DFC system as its underlying rails, either by serving as as a proxy wallet-manager for consumers, or as a genuine payments intermediary by holding DFC balances as collateral against account balances on a 1:1 basis, similar to how MMFs do with treasury securities, or how mobile money systems use ring-fenced deposit (or CB) accounts or trusts to ensure safety of their mobile money liabilities, without incurring significant intermediary risk (depending on the settlement technology, etc).

      The former approach, which I think is cleaner/simpler, is similar to if you delegated to your bank the authority to open your private safety deposit box, and make transfers of the contents to other safety deposit boxes in the approved banking network. The customer, as opposed to the intermediary, still "owns" the property/file in the box, but all transmission signals are intermediated via the intermediary, with whom the individual is required to have an account.

    3. This exchange, between eCurrency CEO Jonathan Dharmapalan and a representative from US Treasury, may shine a bit more light (beginning around 22:10):

      "“Q: Hi, Loretta Michaels, U.S. Treasury. I have to say, we are very excited by what Jonathan is up to. It's been very clear to those of us at Treasury watching this space that there are two areas that we feel strongly about. One is that digital is clearly the future, but we want currencies to be regulated, and not a wild west of unregulated currencies. And so it was quite timely for us to meet Jonathan and hear his thoughts. So we've been very excited by that.

      And so I had a couple of questions for you. One is – how easy is it for the marketplace to adjust itself to accept digital currency? So i'm thinking merchant acceptance, POS systems, ATMs, et cetera, because even if the regulator could turn it on overnight, there needs to be a marketplace use for it. [...]

      A: This ties a little bit to Victor's question – what is different now to when there was Digicash? What is different now is that we have mobile money, and that the marketplace and the public have accepted this capability, and they are using it in tremendous amounts. They are using it for remittances, they are sending it to their villlages, they are using it to buy bananas. It is being transacted already. So electronic money is working, and working incredibly well.

      What we have done is enabled the central bank to create a digital currency, and do a one-time swap of digital currency for mobile money. By doing that one-time swap, they will insert their legal sovereign currency into the mobile money system. Once it's in there, there is no more privately issued e-tokens flying around. There are only currency units issued by the central bank. So from a public perspective, that are already used to transacting on Paypal, on M-Pesa, on Airtel money, the use case doesn't change. They still do the same things they do. The only difference is the ultimate trust of the instrument they are using comes to them from the central bank.

      So it was a combination of putting this philosophy together with technology that can actually do that one-time swap in the system.”


    4. Thanks Rohan, that answers my question. It definitely qualifies as a robust form of money.

  4. JP how would you incentivize the 'thousands' of nodes worldwide to maintain these ledgers? With the existing known examples, such as Bitcoin, the incentives are 'mining' fees and/or transaction fees. In your case, the mining fees are not feasible, since that amounts to outsourcing money creation to unknown third parties unrelated to the central bank. That leaves the transaction fees. There, again, the existing examples such as Bitcoin demonstrate that this is a tricky balance. If you don't have the mining fees, you have to either see constant price appreciation to make prior fees look bigger or you have to make transaction fees bigger to make the process worthwhile. Price appreciation goes contrary to the 'peg' while transaction fees won't stand agains the current debit swipe fees that we even have now.

    1. I think mining fees are feasible. Each block would result in an automatic generation of a few new Fedcoins for the miner. There are other ways to do it too. Rather than a successfully-mined block providing a reward of new Fedcoins, the central bank could pay the miner directly from its own stock of existing Fedcoins.

    2. JP,

      Thanks again for a great post! (I have noticed you've become recognized as an expert on CBDC, and rightly so.)

      This particular comment of yours left me unsure about how you see Fedcoin. Like currency, it should be recorded on the RHS of CB balance sheet, right? But only while it's in circulation. Any "own stock of Fedcoins", just like in case of currency, would be off balance sheet.

      Putting Fedcoin (new or existing; I don't really see much difference, just like notes can be re-used or destroyed & replaced) into circulation means a credit to the account "Fedcoin in circulation" (or similar). Right?

      My question: Which account will be debited?

    3. Hi Antti,

      The Fedcoin mechanism would work exactly like cash currently does. Central banks don't push cash directly into circulation; they use commercial banks as their distribution agents, the public pulling cash into circulation. When a client requests cash from his/her banker (and say the banker doesn't have any on hand), the banker has to go to the central bank and ask for banknotes. The central bank provides the cash while at the same time debiting the banker's account at the central bank. So in the end allthat is occurring is a reduction in one item on the right hand side of the central bank's balance sheet, the liability side, and an increase in another item on the RHS.

      Same with Fedcoins.

    4. I get that, JP.

      I was referring to this: "Rather than a successfully-mined block providing a reward of new Fedcoins, the central bank could pay the miner directly from its own stock of existing Fedcoins."

      If the central bank will pay the "miner" in this way, it will ultimately be a debit to its equity, through P&L. Or that's how I understand it. Of course, there's no need to pay in Fedcoin should the miner prefer to get his checking account credited instead ("I don't want cash, or e-cash -- just credit my account").

      So I would say it's not comparable to Bitcoin, and I don't see how it would matter whether the CB creates new Fedcoins or uses ones it has created earlier and is in possession of (unless the existing ones contain some kind of memory and so should be re-used?).

      I don't know if you get what I mean. I just feel that we are, and have always been, giving too much attention to money as an object. And that you continue on that path if you differentiate between newly created Fedcoins and existing Fedcoins in the hands of the central bank. What if the central bank destroys all Fedcoins it receives and issues new ones as needed? I don't see any difference.

    5. "...and I don't see how it would matter whether the CB creates new Fedcoins or uses ones it has created earlier and is in possession of."

      Yes, you're right. I had a certain model of Fedcoin in mind, where all the coins are created at once at the outset, i.e. premined. Say like 20 trillion of them. The central bank keeps all of these in its account and never creates new ones, only putting them into circulation as required. But the model could also involve the central bank creating new Fedcoins on an as-required basis.

  5. I'll continue a bit, to give you some background.

    I believe that Bitcoin is money misunderstood. When new Bitcoins are created, it's like making a credit entry without a debit entry. Not so in the case of Fedcoins.

    With currency, central bank reserves or Fedcoins, there always exists a debit balance, a debt, of same magnitude on the LHS of the balance sheet. These debit balances consist mostly of what the Fed calls "Collateral Held against Federal Reserve Notes":


    So, Fedcoin is a credit balance. There exists collateral, debit balances, against it. There's no collateral held against Bitcoins.

    I've discussed this in my comment to Tony Yates here:


    I don't think I'm saying anything you didn't already know. I just don't get why Bitcoin is considered similar to "fiat money". It's not, and the most significant way it differs from it has nothing to do with Bitcoin being more scarce. Bitcoin is a credit without a debit. That makes Bitcoin unsound, at least to an accountant like me.

    1. I agree - there is no ongoing debit obligation to take bitcoins in the future, unlike bank deposits that are created by bank borrowers who are obliged to provide something to deposit holders in the future.

    2. Antti, I agree with pretty much everything you say in your comment.

    3. Good to hear, JP and Dinero! Would you say that there's a general agreement among economists on this matter?

    4. I would say not, there is not general agreement amongst economists as to why and how loan contracts give the corresponding liabilities that are written against them in the monetary system value, what you call here sound vs unsound.

    5. That's what I thought.

      If I read you correctly, you call the credit balances 'liabilities'? Like in "liability of the bank"? That's the conventional way to put it, so I don't blame you. Nevertheless, I have found it useful to adopt the opposite view and the language that goes with it (not least because central bank liabilities are often not recognized as real liabilities; perhaps rightly so).

      An example:

      Let's imagine that the LHS of the Fed's balance sheet consists of an MBS worth 100 and gold stock worth 100 (at market price). On the RHS of the B/S there's "FR notes in circulation" with face value of 200. No equity.

      The MBS is a liability of the mortgagors. The gold is a liability of the Fed. Isn't the gold an asset of the Fed, you might ask? Well, who is the Fed? The Fed doesn't really own anything. If we wind it down, everything in its possession goes to the parties holding credit balances -- in this case the FR note holders. We could say the Fed is liable to look after the gold that ultimately belongs to someone else. (This view finds also support in the birth story of double-entry bookkeeping in Genoa, but I save that story for some other time.)

      So, the LHS of the Fed B/S consists of liabilities/debt/debit balances. The RHS consists of credit balances which can be viewed as their holders' rights to receive something. Naturally, that kind of right has value if its holder can expect to receive something.

      So, let's wind down the Fed.

      To make it easy, it just so happens that someone holding FR notes worth 100 happens to buy the houses of the people behind the MBS (assume loan-to-value 100 %). The note-holder got the houses worth 100; he was holding a right to receive something worth 100, and did receive it. The mortgagors had a liability to give up something worth 100, and so they did. After the deal they sent the notes they had received to the Fed, which wrote off the MBS and burned the notes.

      Next, another person holding the rest of the FR notes walks to the Fed and takes possession of the gold. He had the right (credit balance) to receive something worth 100, and he did. The Fed had the liability (debit balance; gold stock) to give up the gold in its possession should the bank be wound down, and so it did.

      For the system to be sound, the amount of liabilities, or obligations, should match the amount of rights. People shouldn't be entitled to receive more than others are obliged to give. To me that sounds like the most natural thing.

      What do you think?