Digital fiat currency

A recent presentation and blog by David Andolfatto of the St Louis Fed discusses the idea of a government-backed digital currency – exactly like Bitcoin, but issued by the Fed and exchangeable at par with ‘real’ dollars. Although this radically contradicts the political philosophy of Bitcoin, which was meant to wrest control of the money supply from the Fed, it is an attractive idea because it links the stability of fiat currency to the speed and convenience of the Bitcoin technical platform (known as the blockchain). Although the digital US dollar is still only a concept, Ecuador has already launched a government-backed digital currency pegged to the US dollar, which it (reluctantly) uses as its local currency. The Ecuadorian system is not based on the blockchain, and indeed Ecuador has banned Bitcoin and altcoins, but the Philippines are considering issuing a blockchain-based e-peso. The idea of digital fiat currency clearly has legs.

The implications of digital fiat currency are potentially massive, impacting the role of banks, the money supply, and a myriad of other social and technological areas. Without wishing to diminish the importance of those areas I want to focus on the possibility of using digital fiat to implement a domestic real-time payments system. Before jumping to the solution we need to take a quick look at the shortcomings of existing systems.

Settlement risk in traditional systems

ACH systems generally use a two-step process. The first ‘clearing’ step is where payment instructions are exchanged - Bank A sends a message to Bank B asking them to credit a customer account that they hold. The second – and little understood – ‘settlement’ step is where funds are moved between the banks. This is normally done on an aggregated basis (all the individual payment instructions for a bank over a given period are added together), and on a netted basis (if Bank A sent £1m instructions to Bank B, and Bank B sent £1.1m instructions to Bank A, the net obligation is for Bank B to settle £0.1m to Bank A). The actual movement of funds happens by the Central Bank making a transfer between accounts they hold for the two banks – this is known as settling in Central Bank money.

In batch-based payment systems a bank is able to wait until settlement is complete before crediting their customers according to the individual payment instructions. Hence there is no settlement risk – if a bank fails to settle, the other banks simply do not act on the payment instructions sent by the failed bank. However, a real-time payments system must give (near) immediate value to payees, before settlement is carried out. In such a deferred settlement model this leads to a settlement risk – if Bank B has credited £1m to its customer accounts based Bank A’s instructions, but then Bank A fails to meet its settlement obligations, Bank B is left out of pocket. In the post-2008 world, this settlement risk is a significant cause for concern for banks and regulators.

The settlement risk can be mitigated via net sender caps limiting the total liability a bank can incur, collateralisation enforced by the Central Bank, loss sharing agreements and similar mechanisms, but this in turn creates significant barriers to entry to smaller organizations that want to participate in the Faster Payments system. Increasing the frequency of settlement reduces the amount of settlement risk that can accumulate, and the in the extreme case would become an RTGS system where every payment is individually settled in Central Bank money. While this would eliminate the settlement risk, the high volume of ACH payments places very significant technical demands on participating systems, which can cause problems with performance and reliability.

Eliminating settlement risk with digital currency

A Fedcoin-based real-time payments system eliminates the settlement risk by the radical innovation of not needing a separate settlement phase. A commercial bank would buy Fedcoin from the Fed at par, exchanging $1m in ‘real’ dollars for $1m Fedcoin. The Fedcoin would have inherent value and would only exist on the blockchain, so the Fed would not need to track ownership of Fedcoins on its ledgers (just as it does not track the ownership of individual dollar bills). Like Bitcoin, a Fedcoin would be a bearer instrument, so when Bank A sends Bank B a Fedcoin payment, Bank B immediately owns the Fedcoin. The value is transferred immediately so no additional step is needed to settle in Central Bank money, or synchronise with the Fed’s ledger. Any commercial bank would be able to exchange Fedcoin back into ‘real’ dollars at any time via the Fed. Fedcoin eliminates settlement risk by eliminating the settlement step entirely.

Coming back to the use of Fedcoin purely as an inter-bank mechanism, there is wrinkle in the scenario described above. When Bank A’s customer want to send ‘real’ dollars to Bank B’s customer, there would actually be two parts to the transaction. The first part is as today, Bank A debits ‘real’ dollars from their customer’s account and sends a payment instruction telling Bank B to credit their customer’s account. The second part involves Bank A transferring the equivalent dollar value from its Fedcoin account to Bank B’s Fedcoin account. The Fedcoin is only used for inter-bank bilateral settlement rather like a nostro-vostro arrangement, and there is no need to hold Fedcoin accounts for individual customers.

Multiplying message types

At the moment the Bitcoin-esque messages used to initiate blockchain transactions do not carry anywhere near the amount of data needed in a payment instruction, so the two parts of the transaction would require two separate messages, one to carry the payment instruction and the other to execute the blockchain transaction. The blockchain message would carry a unique identifier cross-referenced to the payment instruction. Reconciliation between these two messages would add significant complexity compared with today’s payment systems where a single message serves both purposes, and the payment instruction also serves as the source of data to drive settlement.

Although it seems superficially attractive to combine the two messages, it would not be technically feasible for all of the payment details to be recorded on the distributed ledger, nor would it be desirable from a business point of view because the ledger details would be visible to every participating bank. As well as the messaging that updates the distributed ledger, the point-to-point delivery of payment instructions would still be required as it is today. This may be good news from a compatibility point of view, but does mean that the distributed ledger and supporting systems and protocols would be additional to the existing real-time infrastructure, adding significant cost and complexity. This problem would still exist even if customers had their own Fedcoin accounts and were able to make peer-to-peer payments; the supporting remittance information would still need to be transferred by a different mechanism.

A bigger blockchain

There are also question-marks over the scalability of some technical aspects of Bitcoin. On a peak day the blockchain currently processes only around 100,000 transactions daily, compared with 100 million transactions on a peak day across a domestic clearing systems. The blockchain is already unwieldy at 30Gb in size, and will rapidly become unmanageable if its growth rate increases by three orders of magnitude. A key role in bitcoin is played by ‘miners’ who use special computers to generate Proofs of Work confirming that transactions are valid. They would not be able to work a thousand times faster than today, and the power usage of mining systems - already a cause for concern – would skyrocket. Increasing the average confirmation time might alleviate that problem, but the current 10 minute confirmation delay is already glacially slow compared with immediate payment systems such as UK Faster Payments, with a 15 second confirmation time.

Ripple misses the boat

The Ripple network attempts to overcome some of the blockchain’s performance challenges by abandoning some of its fundamental principles. Instead of Proof of Work, a simpler consensus model is adopted. This is more efficient, so transaction confirmation is achieved more rapidly, but it does require participants to maintain an explicit list of trusted parties. This would not work at web scale with billions of individual users, but is feasible in a community of domestic banks participating in a real-time clearing. At first sight it looks like Ripple might overcome some of the issues associated with the Bitcoin blockchain, so it is interesting to review Ripple’s reaction to the Fedcoin proposal.

In a comment on Andofatto’s blog, Ripple’s Phil Rapoport proposes a scheme using something he calls ‘Fed-USD’. It is not made clear what a Fed-USD is, but the description makes clear that it is not a true digital currency. Rapoport says that a bank with $1m on deposit with the Fed could receive $1m in Fed-USD on Ripple. Banks could exchange Fed-USD in real-time via the Ripple ledger, which he calls ‘good funds’, implying no settlement risk. However, Rapoport goes on to state that the existing Fed ledger systems ”can issue a call into the Ripple ledger to see what transactions occurred over the night/weekend and make changes between reserve accounts in the legacy systems to reflect the activity on Ripple.” This implies that unlike a true digital currency, a Fed-USD transaction does not actually transfer value. In other words, the Ripple ledger is only recording settlement obligations. Settlement in Central Bank money is achieved by debiting and crediting the reserve accounts at the Fed, which is the same settlement mechanism used by today’s real-time payment systems. This proposal introduces a distributed ledger, which for reasons discussed above would be additional to the central one, but offers no obvious benefits.

Fedcoin for Faster Payments falls short – for now

The conclusion is that a digital fiat currency such as Fedcoin can be used to eliminate settlement, and hence settlement risk, in a real-time domestic payment system, but doing so requires significant additional technical infrastructure compared with current systems, increasing cost and complexity. There are also questions around the feasibility of scaling blockchain technologies by orders of magnitude. Given the settlement risk mitigation measures already in place in today’s payment systems, it is questionable whether the additional costs would be justified by the benefits. Using blockchain-like technology without a digital fiat currency incurs similar costs but no obvious benefits. Cryptocurrency is a rapidly developing area, and things may change overnight, but for now the use of digital fiat currency to settle a real-time domestic payments system does not look like an attractive option.