31/03/2020 by Dominic Hobson 1 Comment
Blockchain Clearing and Settlement for the Real World
L4S is a start-up led by people who know that a successful blockchain revolution in securities clearing and settlement cannot ignore either the technical limitations of the technology or the prevailing dispensation of service providers, market infrastructures, transaction processing technologies and accounting methodologies. If their approach catches on, blockchain in securities services could just be in the foothills of the Slope of Enlightenment.
Nobody knows how much the various parts of the securities services industry have sunk into blockchain or distributed ledger technology (DLT) experiments, proofs-of-concept and pilots since 2016, but it might easily be as much as US$2 billion. And, so far, they have next to nothing to show for it. Paul Dowding, co-founder and head of design at L4S Corporation, has used his understanding of the limitations of current DLT designs to re-think and re-design how DLT can be applied to the securities services industry.
“Name a commercially viable implementation of blockchain technology outside crypto-currencies,” he says. “There isn’t one.” Although multiple firms have applied DLT to securities services workflows successfully, none has succeeded in processing significant volumes at high speed while retaining the core features – mining and consensus – of classic blockchain technologies.
R3, Hyperledger, Ethereum and Digital Asset have all conceded that scalability is a DLT challenge they have yet to solve. “Tellingly, while crypto-currency exchanges record crypto-currency transactions on each of the currencies’ respective blockchains, they use conventional ledgers for their own records and accounting,” notes Dowding.
Technical limits of blockchain technology
He adds that the search for consensus is a greater challenge for conventional blockchain technologies that the costs of mining. “Mining presents challenges of its own, but the real problem is that consensus algorithms by definition exponentially increase the computational demands as they get bigger,” says Dowding. “The bigger they get, the slower they get.”
Exponential is the right term. Simply adding transactions to a processing machine means the volume scales linearly, while voting to achieve consensus on the transactions of every other participant as well means the volume scales exponentially.
A secondary effect of the time- and energy-intensive search for a consensus on valid transactions is a degree of arbitrariness over which transactions settle first, because the sequence is governed by the votes of all other participants in the network. The inevitable solution – breaking transactions into smaller sub-sets that can be computed more quickly – does not achieve scalability. It simply proves that scalability is impossible with unreconstructed DLT.
Worse, smaller groups of transactions – what DLT advocates call “shards” or “clusters” - are more easily subverted by bad actors, so security is weakened as well. “Although distributed ledgers are very hard to corrupt, they can be corrupted, and the Ethereum hard fork is proof of that,” says Dowding. “If even one of these `shards’ or `clusters’ starts to get corrupted, it throws into question the other ledgers because you can’t prove your records are valid from the date of the hack to today.”
Legacy systems cannot be wished away
These problems – lack of scalability, inability to handle volume, and vulnerability to hacking – are bad enough. But there is another disincentive to adopting DLT. This is the lack of what Dowding calls “backward compatibility,” or the inability of DLT to work with legacy systems. There is never going to be the DLT equivalent of a Big Bang, in which every market participant and every market in every asset class and geography switches overnight from legacy systems to DLT.
So any viable DLT system must not only be scalable, indifferent to transaction volumes and secure from cyber-attack, but able to work with existing cash payment and securities custody systems and inter-operate with legacy market processes and infrastructures. It is a challenging list of demands, but Paul Dowding maintains that L4S has solved them all.
Consistency instead of consensus
The firm solves the time and energy problems of searching for consensus by adopting a different approach to validating data shared across multiple computers in a network. It draws on the research-backed idea of “conflict-free replicated data types” that guarantee “strong eventual consistency in distributed systems.”
Though the jargon is not inviting, the idea at the core of this approach is simple: the members of a network can be confident that their copy of the data is accurate (that is to say, there are no conflicts between the copies) by updating and validating the ledger from the transactions alone.
L4S extended and enhanced these properties by adding a means of communicating “completeness” for every member without synchronisation or reconciliation of databases or (as in the case of DLT) voting to achieve a consensus that blocks of transactions are valid.
Networks which adopt this approach achieve “consistency” and “completeness” – that is to say, every member of the network knows they have got a complete and up-to-date copy of all the transactions between members of the network and can identify and rectify missing transactions individually.
The sole purpose of synchronisation, like reconciliation in an analogue system, is to prove that all copies of the ledger are consistent. By skipping the need to synchronise or reconcile or vote to achieve consensus every time the ledger is updated in this way, the L4S DLT network saves time and adds processing capacity.
An “airlock” allows present and future worlds to inter-operate
To solve “backward compatibility” and inter-operability, L4S has introduced what it calls an “airlock.” The “airlock” is a series of matched transactions between the payments bank where fiat currency is held or the custodian bank or central securities depository (CSD) where securities are held and a so-called "depository node” on the DLT network. It allows an existing quantity of fiat currency or securities held outside the DLT network to be represented on the network in digital form without any risk of duplication, not unlike the assets held in escrow that back a Stablecoin.
The “airlock” protects the integrity of issue while allowing digital versions of fiat currency or securities held outside the network to be transferred between accounts on the network. It is, in effect, a systematic form of tokenisation of off-network assets. It allows payments to be made and securities to be delivered on the network, irrespective of whether the assets are issued and held on the network or not. The downside is that delivery of cash or securities to accounts off the network still has to take place via existing cash payments and securities clearing and settlement systems.
Mark Green, CEO of L4S Corporation, sees this clumsiness as a temporary solution, though in this case pro tempore means until all assets – including fiat currency – are issued on to the network, or at least a network that users of the L4S network can access directly. “The future state is to issue, say, equities as digital assets within the network, with no countervailing shares held by a depository bank or CSD off the network,” says Green. In that future state, the depository node will do no more than record the value of assets held on other networks and facilitate transfers of assets between them.
To accelerate progress towards that nirvana, L4S offers to recreate the complete transaction histories of a client, going back decades if necessary. This would enable a user to decommission their existing systems and clear the legacy and inter-operability obstacles completely. However, Green does not expect many clients to choose this radical option. “We expect clients to go slow, and continue to use their existing systems, and interact with existing market infrastructures as they do now,” he says.
This measured approach is most obvious in asset servicing. Until assets are issued digitally on the network, L4S will record asset servicing transactions in the network, but these transactions will reflect the underlying omnibus or segregated account processing of income, corporate action notifications and instructions and the exercise of voting rights by the custodian banks and CSDs.
“The really interesting future state is to have the issuing companies or their registrars or transfer agents as nodes on the network, so that they can deposit dividends or interest payments or rights directly into the accounts of their investors,” says Dowding. “We know what the future looks like, but we have also got to work with today’s world.”
No need to overturn existing accounting systems
This realisation of the need to work within the constraints of existing systems extends to accounting as well as transaction processing systems and to existing methods of issuing and holding cash and securities. This too marks a departure from conventional blockchain thinking.
One of the most striking claims of the entire blockchain revolution is that it replaces double entry book-keeping with single entry book-keeping and provides verification of ledgers without the need for an independent auditor through the creation of a joint ledger accessible by all parties to a transaction.
Unlike traditional accounting techniques, which proceed from a starting balance to an end-balance and analyse separately how transactions lead from one to the other, a classic blockchain-based network derives balances simultaneously from transactions. In fact, this is what causes the retardation of transaction processing and demands large, ever-increasing amounts of active computer memory.
It also means balances can never be stated without reference to the entire string of previous transactions. This in turn means that the balance is out of date as soon as it is stated, because it takes so long to run the calculation that new transactions have overtaken it by the time it is completed.
“I don’t know how people are reconciling to distributed ledgers and meeting auditing standards, argues Dowding. “I haven’t seen any GAAP-approved outputs from these DLT networks. You have to feed it into a conventional accounting system and reconcile it back to the distributed ledger.”
Traditional accounting techniques do not run into this difficulty, because they take the starting balance, add transactions to it, and calculate the end-balance. Transactions and assets remain separate, as they do in the traditional profit-and-loss and balance sheet accounts. This obviates the associated problems of speed and memory too, because it allows past transactions from prior periods to be archived.
L4S has revived this traditional approach for its version of clearing and settlement on a DLT platform. “Because we have maintained the separation between the transactions and the assets, we can now reflect all types of assets on the ledger, including payables and receivables, and loans and accruals,” says Dowding. “That makes it very easy for the banks to absorb the data because we are not changing the format and order.”
Accounting as we know it, but in real-time
The L4S DLT system treats all transactions as conventional payables and receivables and delivers real-time updates to the general ledger, continuously updating the balance sheet in real-time rather than taking periodic snapshots. This allows users to adopt L4S without having to reinvent their accounting systems, because it makes use of the same concepts, just in real-time.
“It’s a dynamic zero-sum game in which transactions continually update who owns what and who owes what in near real-time, where the transactions and entire ledger are validated, posted and reconciled on an event-driven basis,” explains Dowding. “Our premise is that the Internet works because it is a consistent protocol that interacts with the computers and other devices that people are using. The users do not care how it works. In the same way, participants in the securities markets don’t want to have to change their order management and accounting systems or translate their processes. They just want to interact with the technology.”
But insulating users from technical change still allows counterparty and systemic risk to be managed in new ways. “From a risk perspective, you can see the balance sheet and leverage and duration of liabilities of a counterparty in real-time,” explains Dowding. “In the financial crisis, for example, Lehman counterparties would have seen the Lehman balance sheet extending from a leverage perspective and going out in terms of duration, and people would have just slowly backed away rather than let them go off a cliff. That’s one of the things we think the regulators will like about L4S. They can aggregate total market numbers. They could be a monitoring node, and see the level of activity, the aggregate leverage, the exposures.”
Real-time settlement no longer depends on pre-funded accounts
Dowding also contends that retaining a familiar accounting methodology solves one of the biggest obstacles to real-time settlement. This is the need to ensure the necessary cash or securities are always available – in other words, that accounts are pre-funded, and not dependent on the receipt of funds or securities from another transaction, or the ability to borrow them whenever they are needed. Instead, explains Dowding, L4S DLT network users will be able to treat all transactions as intra-day payables and receivables, settled in order of value or importance.
“The reason people think pre-funding is a cost of real-time settlement is because they are still thinking in terms of overnight funding and that anything they borrow today will settle sometime today, possibly as late as 4 o’clock,” says Dowding. “In future, settlement is going to work in the same way that trading works. Just as a trader says the securities are available in this amount at this price for the next five minutes, the operations people are going to say, `I’m happy to settle this transaction at this price, but if you don’t settle in the next five minutes, I’m going to charge you `x’ per cent in interest for every minute after that.’”
Of course, the obstacles to a marketplace in which the costs of settlement between pairs of counterparties are governed by the price of an intra-day overdraft are not trivial. Treasury desks, and financing desks, will have to stop working to a T+2 settlement timetable in which trades settle in overnight batches and start working in real-time. Which means transitioning away from current clearing, settlement and accounting systems. “You can only do this if your technology supports real-time, irrevocable transactions,” admits Green.
But Dowding argues that the transition is less disruptive than it sounds. In his view, the middle and back office merely require access to the data captured by the real-time position-monitoring and profit-and-loss calculation technologies of the front office. In other words, to settle transactions executed by traders, the treasury desk will source the cash or securities from internal accounts in near real-time or the financing desk will borrow the cash or securities in near real-time.
Netting and real-time settlement are not binary choices
Dowding thinks certain pairs of counterparties will be comfortable enough with their reciprocal credit risks to net trades off at the end of the day – albeit within intra-day credit limits - rather than settle in real-time.
“The choice is not binary,” he explains. “It’s not real-time gross settlement or net settlement, with nothing in between. In fact, you will be able to net by counterparty, as and when you want to. You can settle every transaction in real-time, or you can allow the exposures to expand for five or ten or 15 minutes or an hour, and then net them. What you choose to do can vary by counterparty, with credit and risk officers adjusting exposures in real-time.”
Dowding reckons the balance sheet savings from near-real-time settlement will be attractive to the sell-side, while the elimination of counterparty risk will please the buy-side and reassure regulators.
Existing providers can survive but only if they adapt
Existing service providers, such as custodian banks, CSDs and central counterparty clearing houses (CCPs) will wonder what use this dispensation will mean for them. Dowding thinks netting at CCPs will continue – provided counterparties choose to novate their trades to them - but is less optimistic about the future of conventional custody services provided by banks and CSDs.
He believes that L4S-style DLT networks will disintermediate custodians and CSDs from transaction processing but expects them to retain a role in client on-boarding and transaction reporting. “People talk about these Utopian peer-to-peer networks,” says Dowding. “They forget you have to complete Know Your Client (KYC) and anti-money laundering (AML) checks to ensure your client is not on the Office of Foreign Assets Control (OFAC) lists.”
He predicts that a “transaction protocol” will emerge akin to the TCP/IP Internet Protocol that governs exchanges of information across the Internet, allowing equities, bonds, futures, options and over-the-counter (OTC) derivatives to be cleared and settled by a single process. Custodians will effectively act as conduits for clients using the protocol to settle transactions and hold assets.
“This is not going to be a faster, better, cheaper version of what custodians do today,” he says. “Custodian banks will still exist, but they will be transaction hosting services rather than the entities that they are today.”
This transition to a new role for custodians reflects the fact that the L4S approach to its transaction protocol is a radical departure from similar DLT designs. “Despite the apparent variety, all current DLT solutions rely on the untenable processing of competing validators or network consensus,” explains Dowding. “The L4S DLT protocol is as different to consensus-driven DLT solutions as the TCP/IP protocol was to Wide Area Networks (WANS).”
Dowding believes custodians and CSDs have no choice about the need to adapt anyway, because new entrants from outside the industry will supplant them if they do not change. “DLT lowers the cost of entry,” he says. “There are a lot of networked companies – the Walmarts, the Verizons, the Amazons – that, but for a banking licence or a brokerage licence, could jump in and offer financial services. With real-time and near real-time settlement, the applications can easily be built to represent securities transactions back to the client and these types of companies can expand the use of this technology into other areas for their clients, such as retail purchases, supply chains and the Internet.
Written by Dominic Hobson - May 2020