The trade-off between shorter settlement times and multilateral netting benefits in deferred net settlement

Dennis McLaughlin

1 Introduction

This paper attempts to estimate the benefit from multilateral netting between market participants over an intraday settlement window that is shorter than the length of the trade day. Our approach shows that, as settlement windows get shorter, the multilateral benefit decreases accordingly. In the limit, no benefit remains and in effect there is real-time gross settlement (RTGS). The paper aims to answer the question of how short the settlement window can be in a deferred net settlement (DNS) system without losing any multilateral netting benefits.

When the model is fitted to National Securities Clearing Corporation (NSCC) statistics covering the equities market in the United States, it shows that, for the network as a whole, in theory there is no material loss of the multilateral netting benefit for a settlement window of at least one hour. Note that the benefit today (in a $T+2$ world), measured as the ratio of net to gross transaction values, is well over 90%. Our approach also demonstrates that the benefits of multilateral netting accrue very quickly during the window, and that the hubs (where most exposures sit) can be expected to achieve their maximum possible netting benefit within the first 30 minutes. In fact, after just 10 minutes, the hubs can be expected to reach 90% of their maximum benefit. The approach also demonstrates the complete loss of any multilateral netting benefit for atomic settlement, as expected. The methodology can be applied easily to any DNS system.

The paper also discusses the practical and theoretical challenges of achieving a one-hour settlement window in DNS, and the extent to which distributed ledger technology built on blockchain might be a feasible solution. The major obstacles to overcome when using a distributed ledger technology approach will be the creation of very fast interoperability between the underlying blockchains and legacy systems, the speed of the underlying consensus algorithms, and the speed in building out a connected system of counterparties with critical mass.

This paper does not comment on how (or to what extent) the loss of multilateral benefits in DNS could be addressed using an RTGS system, perhaps supplemented by a liquidity savings mechanism to achieve the same settlement results. Nor indeed does the paper comment on which system is the better one; that is, whether the higher credit risk and lower liquidity needs in a DNS system are better than the higher liquidity needs and lower credit risk in an RTGS system. Indeed, there are large-value RTGS systems in operation today that achieve something close to instantaneous settlement (settlement in minutes) and avoid DNS, as many central banks have successfully navigated the move away from DNS to RTGS (see, for example, Byck and Heijmans 2020). A discussion of how this might be accomplished in the equities markets is beyond the scope of this paper and will be taken up elsewhere.

2 Context

There are currently major efforts underway globally to shorten the settlement window in the securities markets. In the United States the underlying motivation comes ultimately from the Bachmann Report issued in May 1992, which is based on the philosophy of “time equals risk” and thus “less time between a transaction and its completion reduces risk” (Bachmann Task Force 1992). The topic of quicker settlement was prioritized for urgent consideration after the market volatility following the outbreak of the Covid-19 pandemic and the heightened volatility around certain “meme” stocks (US Securities and Exchange Commission 2022a). The situation globally at the time of writing is as follows.

United States.

After extensive market consultation, the Securities and Exchange Commission has ruled that the securities settlement window will be shortened from $T+2$ to $T+1$ , with the compliance date set for May 28, 2024 (US Securities and Exchange Commission 2023).

Canada.

The Canadian Capital Markets Association has chosen to follow the United States, and the move to $T+1$ will be implemented at the same time as in the United States in 2024 (Canadian Capital Markets Association 2023).

India.

On January 27, 2023, the Securities and Exchange Board of India announced that the Indian market had already completed its yearlong transition to a $T+1$ settlement cycle (Mathew 2023).

United Kingdom.

HM Treasury has created an industry task force to study the case for moving to a $T+1$ standard settlement cycle, with a full report and recommendations due in December 2024 (HM Treasury 2022).

European Union.

On October 5, 2023, the European Securities and Markets Authority launched a “call for evidence on shortening the settlement cycle”, with a closing date for comments of December 15, 2023 (European Securities and Markets Authority 2023).

A shorter settlement cycle has the following obvious benefits.

Reduced counterparty risk.

As the time between trade execution and settlement in full is shortened, there is less time for a counterparty to default and less potential movement in the price of the security, which should therefore result in smaller margins overall.

Capital efficiency.

Resources freed up from lower margins held over the shorter settlement window can be redeployed sooner, driving more efficient use of capital.

Improved liquidity.

Having less funds tied up, and for shorter periods, should ultimately increase trading volumes and liquidity in the market.

To understand just how material multilateral netting is today, consider the statistics revealed by the chief executive officer of the Depository Trust and Clearing Corporation (DTCC) in congressional testimony (Bodson 2021). On a normal day in 2020, the NSCC processed about 200 million transaction sides, and the daily value of trade sides destined for netting on $T+2$ was nearly USD2 trillion. These were trades between the 155 full-service members of the NSCC. These were then netted down at the end of $T+2$ , so that only about USD35 billion ended up physically settling, with 1 million securities movements. The benefit of netting at the end of the settlement window results in about a 98% reduction in physical settlements and greatly reduces the associated operational risks of having to physically settle a very large number of trades. This entire process was all supported by capital of only USD6–8 billion (Pozmanter 2021). In other words, capital of close to USD8 billion could support gross transactions of almost USD2 trillion over the settlement window, before this was netted down at the end of the settlement window to about USD35 billion in physical cash settlements.

The striking 98% net-down at the end of the settlement window is the direct result of the existence in the network of a clearing function – or central counterparty (CCP) – that stands between the counterparties in bilateral trading. The CCP is the “buyer to all sellers” and “seller to all buyers” in this market structure, and it allows all participants to net down all their trade exposures to each other at the end of the window, resulting in only a positive or negative net exposure to the CCP. This is the process of multilateral netting.

For a given cohort of trades executed between all market participants on $T+0$ (the trade date), the aggregate risk can be expected to grow with the square root of time over the following two days until settlement at the end of day $T+2$ . Since $\sqrt{2}\approx 1.41$ , reducing the $T+2$ settlement window to $T+1$ should free up over 30% of the risk margins held for this cohort, in agreement with Depository Trust and Clearing Corporation (2022). Moreover, on any specific day, there are two cohorts of trades to consider, one settling at the end of today and one settling tomorrow. This means that the total potential margin savings is twice the 30% margin savings on any one cohort.

Ignoring for now any practical difficulties of reengineering operational processes, a further reduction of the settlement window to the end of the trade date can be expected to reduce risk margins even further, by about another 30% for each cohort. Though now there is only one cohort for settlement each day, so there is no need for margins to cover a second cohort. This means that the potential saving from moving from $T+1$ to the end of the trade date is far more than 30%.

Gross-to-net ratios of more than 90% are common in major markets globally, showing just how efficient and beneficial the settlement process today really is. For example, TARGET2-Securities (T2S) has reported close to 94% settlement efficiency (European Central Bank 2023), while CLS reported netting efficiency around 99% from multilateral netting in their peer-to-peer foreign exchange (FX) settlements (CLS Group 2022).

Given these very attractive benefits, it is natural to ask: why not continue to reduce the settlement window in the securities markets further below $T+1$ to $T+0$ (same-day settlement), or even further to near-instant settlement?

The US Securities and Exchange Commission has also asked a similar question and invited comment: “Is it possible to shorten the settlement cycle in the US markets to $T+0$ and retain multilateral netting? If so, what is the earliest time on $T+0$ that market participants could be prepared to settle their trades without eliminating multilateral netting …?” (US Securities and Exchange Commission (2022a, Q. 73, p. 119)) The remainder of this paper is an attempt to shed some light on this question.

This problem is also relevant to many very important industry settlement processes worldwide, particularly as they consider migrating from DNS to RTGS. The Payments Canada situation, for example, is described in detail by Byck and Heijmans (2020).

3 Some critical challenges for shorter settlement

The practical difficulties that must be overcome in reducing settlement times in the equities DNS system to $T+1$ have been identified by the industry in many comment letters and were carefully considered prior to mandating the final rule change in the United States (see for example, US Securities and Exchange Commission 2023, 2022b). Similar consultation periods are underway in the European Union and United Kingdom as these jurisdictions consider synchronizing with the United States.

The critical issues identified by the industry so far are the following.

Coordination across different time zones.

Capital markets are global and changes in one major market will inevitably impact counterparties and investors across the globe. This drives the effort to synchronize settlement times. The greater the time-zone difference between two markets, the harder it will be to navigate the shift to accelerated settlement times – hence the need to coordinate changes across countries, time zones and different legal or regulatory jurisdictions. It is very difficult to successfully coordinate all these moving parts.

Compressed post-trade activities.

Moving from $T+2$ to $T+1$ cuts the time taken to execute today’s post-trade processes by 50% or more depending on the market. This means, for example, less time to engage with counterparties, resolve any breaks or mismatches and to complete reconciliation activities, so failures may be expected to increase, at least in the short term.

Cross-currency issues.

•
Adopting $T+1$ raises the issue that transaction funding depending on FX settlement may not occur on time. The scale of the issue is not yet fully understood, but according to the US Treasury, foreign investment in US securities is significant, in that about 20% of US securities (around USD25 trillion) are foreign owned (Global Financial Markets Association 2023).
•
Any trade that needs to be settled in a foreign currency could create a challenge, as there will be a mismatch between the $T+1$ settlement window in the United States and the $T+2$ settlement window for CLS. So, if a trade is executed by a foreign counterparty on $T+0$ in the United States, the FX trade will need to have already been booked with CLS for settlement to complete on time, or else it would need to be done outside CLS (likely through a forward or future). This is likely to raise funding costs and increase risk.

Additional challenges faced by investment managers.

•
The challenges posed by mutual funds mean that in the United States, for example, an industry solution is required to achieve $T+1$ . The DTCC will need to complete a systematic update of the Funds/SERV system for domestic securities that will shorten the settlement cycle from $T+2$ to $T+1$ , effective May 28, 2024.
•
Since funds are generally valued after the close of the markets, it will be difficult to bring the settlement cycle down below the end of the trade date without a further fundamental restructuring in the industry.

There are many other issues, such as corporate actions, mergers and acquisitions, and dividends, as well as interactions between the securities markets and various other markets, such as those for options, repos or stock lending – all of which need reengineering. The industry in the United States has been working to resolve all these practical challenges and get to $T+1$ settlement within the existing DNS framework without materially losing the benefit of multilateral netting described above. Further shortening the settlement window to close to $T+0$ will likely not be achievable without further significant changes across the industry. In fact, the Securities Industry and Financial Markets Association found that “the law of diminishing returns applies: shortening the settlement cycle beyond one day embeds more risk without creating additional benefits available for widespread adoption across the industry” (Bentsen 2022). Moving the settlement window to before the end of the trade date in the current operating environment is akin to “breaking the sound barrier” in the industry.

4 Estimating the multilateral netting benefit over shorter settlement windows

Putting these practical considerations to one side, it is natural to ask whether there is some theoretical limit to how small the settlement widow could be. We have seen above that it is easy to estimate the margin benefits of going from $T+2$ to $T+1$ to $T+0$ using the square-root-of-time rule, essentially because we can estimate the risk change over the next few days for a fixed cohort of trades starting at the end of the trade date $T+0$ . We can also broadly identify those operational processes within the current environment that we need to bypass if we are still to retain significant multilateral netting benefits for a window beyond the trade date $T+0$ .

Any potential benefits of a further reduction in settlement times to before the end of the trade date are more difficult to estimate, and any solution will need to model the trading network prior to the end of the settlement window. We proceed as follows.

(1)
We recall the seminal paper by Duffie and Zhu (2011), which outlines a framework to estimate the benefit from multilateral netting at the end of the settlement window (which here we assume is the end of the trade date).
(2)
We next describe the Barabási–Albert model (for more details, see Barabási and Pósfai (2016)), which enables us to view the network at an earlier point in the day and estimate node connectivity at that point, and hence the expected multilateral netting benefit to any node at that point too.
(3)
Using (1) and (2), we can then interpolate backward in time to derive the dependence of the maximum net-to-gross multilateral netting benefit over the length of any settlement window prior to end of the trade date.

First, we recall the Duffie–Zhu framework (Duffie and Zhu 2011).

Assume there are $N$ homogeneous counterparties who trade with each other in the network over the course of a day. Let $X_{ij}$ be the amount that $j$ owes to $i$ . If $X_{ij}$ is positive, then the default of party $j$ will mean that $i$ suffers a loss, and otherwise there is no loss to $i$ . Hence, the exposure of party $i$ to counterparty $j$ is $\max\{X_{ij},0\}$ .

Assume that each node has traded with all other nodes over the course of the day, so the connectivity of each node is $N-1$ and the network is completely connected at the end of the window. This is certainly a reasonable assumption in the NSCC example mentioned above, as on a normal trade date there are approximately 100 million trades before the CCP steps in at the end of the day.

Duffie and Zhu further assume that each $X_{ij}$ is independent and identically normally distributed with mean $0$ and variance $\sigma^{2}$ .

They then calculate that in this framework, and before the introduction of a central clearing node, the expected exposure of node $i$ to all the other nodes of the network is

$(N-1)\sigma(2\pi)^{-1/2}.$

(4.1)

This can be pedantically expressed as the expected connectivity of node $i$ times $\sigma(2\pi)^{-1/2}$ .

Next, they introduce a central counterparty into the network and recalculate the exposure of node $i$ to be

$\bigg{[}\frac{N-1}{2\pi}\bigg{]}^{1/2}\sigma.$

(4.2)

This formula is the expected exposure of node $i$ , which is calculated by integrating the quantity $\max\{X_{i1}+X_{i2}+\cdots+X_{iN-1},0\}$ against the normal density function, reflecting the multilateral netting of all network exposures after novation to the CCP. The calculation uses the fact that the sum of the $N$ normal random variables is itself normal with mean $0$ and variance $N\sigma^{2}$ .

The benefit of multilateral netting through the operation of a central clearing node within the network can then be approximated at the end of the horizon by dividing the post-CCP exposure expression (4.2) by the pre-CCP exposure expression (4.1) and cancelling terms to arrive at the following expression for one minus the net-to-gross exposure:

$q_{i}(1):=1-\frac{1}{\sqrt{N-1}}$

(multilateral benefit of netting at time 1).

The objective now is to estimate $q_{i}(t)$ at times $t<1$ .

At this stage, some preliminary observations can be made on the shape of the curve $q(t)$ , where $t$ is the length of the settlement window.

•
Note that, in the derivation of $q$ , the contributions from the exposures at each link cancel and the only contribution left is from the connectivity of the node degree.
•
As $t\to 0$ the time intervals become so small that there are only a few trades possible, so that, as you approach the first trade between two nodes, $1-q(t)$ must approach $0$ (ie, there is no benefit from multilateral netting over a small enough timescale). This case is sometimes known as “atomic settlement”.
•
At the end of the day, all $N$ parties will have traded with each other, so as $t\to 1$ , the netting benefit becomes closer to 100%. In the NSCC example, the benefit is about 98%.
•
Hence, the curve $t\to q(t)$ has the overall trend of sloping upward with time, starting from $0$ at $t=0$ and finishing somewhere near $1$ at $t=1$ .

We now extend the Duffie–Zhu framework to the network $S(t)$ at any time $t<1$ , as follows.

(1)
Let $k(t)_{i}$ be the degree of node $i$ at time $t$ (ie, the number of links to other nodes at time $t$ ). At an earlier time $t$ , the network will not yet have grown to be complete, so there may be nodes at time $t$ with degrees less than $N-1$ .
(2)
Now assume that all nodes at time $t$ have exposures to the other nodes, which are identically normally distributed and independent with mean 0 and the same variance $\sigma(t)^{2}$ , possibly different from the variance at $t=1$ . The variance $\sigma(t)^{2}$ can be interpreted as the average variance among active nodes at time $t$ and will of course evolve with time. Note that the older nodes are expected to have traded more than the newer nodes by time $t$ , so using the average variance allows us to stay within the homogeneous framework.

This extends the Duffie–Zhu framework in that, a priori, we do not know what the distribution of exposures between two parties will be, only that it is zero at the beginning of the window and symmetric (equally likely to be long or short), and that trades between different parties are independent. Hence, the assumption of independent and identical normal distribution is appropriate, and the risk exposure measure used will scale with the variance. Of course, this approach can be refined with an ex post analysis of the empirical distributions observed, but that is unnecessary for the immediate purpose and would require a vast amount of (unavailable) data on each node: the multilateral netting benefit we are after here is a relative quantity, as it is captured by a net-to-gross ratio, with the exposure distribution appearing in both the numerator and denominator of this ratio.

We can now derive the multilateral benefit $q_{i}(t)$ for node $i$ in exactly the same way as for $t=1$ above, and since again the contributions from the standard deviations at the nodes cancel, this leaves the following expression for the multilateral netting benefit at time $t$ :

$q_{i}(t):=1-\frac{1}{\sqrt{k(t)_{i}}}.$

Note that the benefit of multilateral netting for a node with a higher degree is greater than for a node with a lower degree.

We look for a formula for $k(t)_{i}$ under two basic assumptions.

Assumption 4.1 (No node growth).

The number of potential nodes $N$ remains constant and the network only develops by adding links between these nodes at a constant rate.

Assumption 4.2 (The Matthew principle).

This principle is named for a verse in the Gospel of Matthew: “For whosoever hath, to him shall be given, and he shall have more abundance: but whosoever hath not, from him shall be taken away even that he hath” (Matthew 13:12, Authorized (King James) Version). This is often paraphrased as “the rich get richer and the poor get poorer”.

In the present context, the Matthew principle translates into the law of preferential attachment, whereby new nodes tend to link to the more connected nodes. This contrasts with nodes in random networks that randomly choose their linking partners.

These assumptions are realistic in the context of a trading network.

•
Assumption 4.1 reflects the fact that there is a finite network of names that trade which each other, and that it is very difficult for a new name to enter the network, as participants would have to subject it to a raft of checks, tests and credit assessments before proceeding to trade with it. Hence, it is reasonable to assume that the universe of possible names is constant to begin with and that all potential trades are between these names, and thus that the network grows by forming more links between them.
•
Assumption 4.2 is also realistic in that there are certain large names that participants tend to trade with. More and more trades accrue to these names by virtue of their size, and they emerge as hubs.

This is precisely the Barabási–Albert model, designed originally to isolate those features that give rise generally to scale-free networks. The key points from its analysis in Barabási and Pósfai (2016, Chapter 5, Model B) are as follows.

Mathematically, if we approximate the degree $k(t)_{i}$ with a continuous real variable, representing its expectation value over many realizations of the evolution of the network, then the rate at which an existing node $i$ acquires links as the result of new nodes connecting to it is

$\frac{\mathrm{d}k_{i}}{\mathrm{d}t}=m\frac{k_{i}}{\sum^{\prime}_{j}k_{j}},$

where the coefficient $m$ specifies that each new node arrives at a constant rate with $m$ links, and where $\sum^{\prime}_{j}$ is the sum over all nodes other than node $i$ , and so must equal $2mt-m$ .

This model is then solved to give a formula for the degree of node $i$ at time $t$ :

$k(t)_{i}=m\bigg{(}\frac{t}{t_{i}}\bigg{)}^{\!1/2},$

for a node $i$ that joins the network at time $t_{i}<t$ .

Note that

•
all node degrees follow this law in the network;
•
the earlier that node $i$ joins the network, the larger $k(t)_{i}$ will be;
•
this network grows at a constant rate into a completely connected network in which all nodes have complete degree $N-1$ as in the Duffie–Zhu framework (Duffie and Zhu 2011); and
•
the resulting network is not scale-free for large time $t$ , although it does follow a power law in the early stages of its evolution.

If we choose node $i$ to be the earliest node in the formation of the network, which activates at time $t_{i}$ , then the expected degree of this node at any later time $t$ will be the highest among all nodes, and so we arrive at the formula for the largest potential netting benefit of any node in the network at time $t$ :

$q_{i}(t):=1-\frac{1}{\sqrt{k(t)_{i}}},$

(4.3)

where $k(t)_{i}=m(t/t_{i})^{1/2}$ .

5 Calibrating the model to NSCC data

The Duffie–Zhu framework was developed for a derivatives portfolio in which the value of each instrument at the outset was zero and the risk exposure was deemed to be the standard deviation. This also captures the essence of other instruments such as stocks, futures and repos, since the mark of the net position of both cash and security is zero at the outset. The risk is then a decline in the value of the other leg relative to the cash leg, since nothing changes hands until the end of the window (eg, there would be no principal risk if counterparties were using a triparty mechanism for bilateral trading).

With 100 million transactions on a normal day between 155 members prior to the NSCC stepping between them, and assuming a 12-hour day (43 200 seconds), this would mean on average 2314 trades per second or 14.9 trades per member per second prior to netting at the end of the one-day period.

If each time step is $1/14.9$ seconds, then $m$ is about $1$ and $t_{1}$ is $1/14.9$ seconds. Using (4.3), it will take at most about 1588 seconds (about 26 minutes) for the oldest node to reach full connectivity with the 154 other nodes.

The fit is shown in Figure 1. The scale of the $x$ -axis is time in minutes and the $y$ -axis is the multilateral benefit achieved by the earliest node at each point $t$ during the day, if the window has length $t$ . This is the relative benefit of multilateral netting over pure bilateral netting due to the central clearing mechanism.

The graph shows the maximum possible multilateral benefit achievable by a node in this model, as it is the benefit associated with the earliest node. The curve acts as an envelope for all the benefits achievable by other nodes.

The most notable feature is just how quickly the maximum possible benefit can be achieved by the earliest node: it takes at most just 26 minutes to reach the maximum possible benefit and at most just 10 minutes to achieve 90% of the possible benefit. After at most 26 minutes, the connectivity of the dominant node will stay at 154 (fully connected) for the rest of the window, so the graph of $q_{i}(t)$ will then flatline at a 92% benefit. Note that this is lower than the actual benefit of more than 98% achieved by the NSCC; that metric, however, was based on US dollar exposures and (as expected) is higher than this homogeneous model’s output due to the presence of large hubs.

In fact, the largest players are very likely to be among the earliest to activate in the network, so the time taken for these hubs to reach full connectivity with the other 154 members will also be very close to the 26 minutes seen for the earliest node, and these hubs can also be expected to reach 90% of the maximum benefit in around 10 minutes.

Since each node follows the same growth law, the last node to which the earliest node links will itself be linked to all other nodes in at most a further 26 minutes. This means that all nodes in the network become completely connected to all other nodes in at most twice 26 minutes (ie, 52 minutes).

A settlement window close to zero (atomic settlement) presents issues since the mechanism breaks down due to a scarcity of active counterparties. However, the graph shows that the netting mechanism begins to take hold rapidly even after only one minute, since by then 138 840 trades would have already occurred in the network at the NSCC.

We summarize this discussion in the following section.

6 Main result

Suppose that $S(t)$ is a large homogeneous trading network of $N$ fixed nodes at time $t$ , which evolve over time by adding links between themselves at a constant rate according to the preferential growth hypothesis (see Assumption 4.2).

If the NSCC statistics are fitted to this model as above, then we have the following results.

•
The network becomes completely connected in just under one hour, and every participant has maximal connectivity after that.
•
If the settlement window is chosen to be at least one hour long (or more accurately at least 52 minutes), then there is no material loss of multilateral netting benefits across the network.
•
The hubs in the network will very likely reach their maximum possible multilateral netting benefit in about 30 minutes and reach 90% of the maximum potential benefit in about 10 minutes.
•
No theoretical multilateral netting benefit is possible for instantaneous settlement, as the settlement window at the atomic level would be $1/14.9$ seconds and $q_{i}$ is zero at that timescale (there are only two active nodes at this scale, so no multilateral netting is possible).

The model used here is also directly applicable to many other large multilateral netting processes currently in use globally and would yield very similar results.

7 Concluding remarks

We have seen in Section 5 that the settlement window in the case of the NSCC could in theory be compressed to one hour without any loss of multilateral netting benefits. Of course, this requires participants to be able to “recycle” their liquidity at the end of each hour over the course of the day. The interesting question now is how a one-hour settlement window could possibly be achieved in practice considering the obstacles identified in Section 3.

Several studies have advanced a possible theoretical solution that is based on distributed ledger technology (DLT) driven by blockchain and a fast underlying consensus algorithm (Feenan et al 2021; Asset Servicing Times 2021). While a detailed discussion is beyond the scope of this paper, this is a very active area, though with many fundamental issues still open (see McLaughlin 2024). Given the results of this paper, there are several comments to highlight in regard to DLT.

The DLT solution and its associated consensus algorithm would need to scale to be viable. In the case of the NSCC statistics, the underlying consensus algorithm would need to validate over 200 million transaction sides over the course of the day: close to 5000 a second between the 155 full members, which then need to be netted at the CCP. This is currently out of reach for DLT, though Ethereum has a road map to significantly speed up validation, and a consensus algorithm based on Ethereum could achieve this soon using “proof of stake” and “sharding”.¹¹ 1 URL: https://ethereum.org/en/roadmap/.

Instantaneous settlement (also known as atomic settlement) is a hot topic in DLT, with some holding it out as the way forward for settlement. Although such a mechanism would reduce capital requirements, the discussion in Section 5 shows that, unfortunately, the benefit of multilateral netting is completely lost at the atomic level and the system is left with the problem that at the instant of the trade both sides must be fully ready and able to transact. One party must have 100% of the cash available and the other side must have the security already in full possession. Positions on both sides must be visible to the other at that instant. This is precisely an RTGS system. A fundamental redesign would then be needed to transition from a DNS system to such an RTGS system, and a lot of planning would therefore be required. Since each side of the transaction would need to be 100% funded at the precise instant of the trade, this would entail a very different market structure, as new trade-financing mechanisms would have to be introduced.

Instantaneous settlement is tantamount to a fully funded payment system, with no credit extension. Monahon (2019) pointed out that such a system had been tried before, by merchant traders in Italy in the 1500s in an attempt to curb credit extension through ledger entries. How exactly this might work in equity markets did not end well, as there was no fully developed RTGS process waiting to replace the DNS process. Instantaneous settlement has really nothing to do with technology, but rather is a consequence of correctly understanding the trade-off between credit extension and liquidity requirements. Indeed, many central banks run very successful RTGS processes today, showing that instantaneous settlement can be done using various liquidity savings mechanisms. This is still very much an open area of current research as more settlement systems around the world migrate to RTGS (Byck and Heijmans 2020).

Nevertheless, there is a lot of value to be gained in the short term from the new technology. As shown in Section 3, there are some very difficult practical obstacles to moving the settlement window to the end of the trade date, never mind down to one hour. The DLT solutions proposed can in principle bypass the legacy infrastructure issues and use a blockchain to track a distributed ledger of relevant transactions over the course of a one-hour window and then net down the trades that have occurred during the window. It is highly unlikely, though, that there will be only one blockchain in the future, as many distinct versions already either exist in production or are in development (see, for example, Kharif 2023).²² 2 See also the Fnality home page: http://www.fnality.org/home. Several bilateral systems are also either in use or in development and can handle targeted intercompany transactions.³³ 3 See the Baton Systems home page: https://batonsystems.com/. The implication is that interoperability between blockchain distributed ledgers and traditional central ledgers will be key to ensuring fast processing times in order to handle quick settlement at the scale needed in major markets. The experience of the Australian Securities Exchange and the failure of the DLT solution there shows how difficult this is to achieve in practice, even with sufficient funding and the support of all parties in the market (Clancy 2022).

The model presented in this paper shows that realizing the multilateral netting benefit over the settlement window is critically dependent on achieving the network effect from connecting more and more counterparties in the market. Getting a critical mass of liquidity in any system that uses this technological solution will be one of the main drivers of success.

Finally, note that this paper does not rule out the possibility of a hybrid settlement system – that is, of supplementing a DNS system with an RTGS system to compensate for the loss of multilateral netting benefits and produce an effective settlement horizon of only a few minutes. This is difficult to engineer but has been done with a lot of work (see, for example, Byck and Heijmans 2020). Ultimately, the regulators and industry must agree that this is the right approach for the particular market in question.

Declaration of interest

The author reports no conflicts of interest. The author alone is responsible for the content and writing of the paper. The paper reflects the opinion of the author, who is responsible for any errors. The author is an independent board director at the Canadian Clearing Corporation and the Canadian Depository for Securities Ltd.

Acknowledgements

The author thanks the anonymous referee for providing some clarifying insights.

References

Asset Servicing Times (2021). $T+\text{2}=\text{antiquated}$ or reliable? Asset Servicing Times, March 3. URL: https://bit.ly/3vulNZh.
Bachmann Task Force (1992). Report of the Bachmann Task Force on clearance and settlement reform in US securities markets. Report, May, US Securities and Exchange Commission, Washington, DC. URL: https://bit.ly/3TVKXde.
Barabási, A.-L., and Pósfai, M. (2016). Network Science. Cambridge University Press.
Bentsen, K. E., Jr. (2022). $T+\text{0}$ ? More risk, fewer benefits. News Article, February 2, Securities Industry and Financial Markets Association, Washington, DC. URL: http://www.sifma.org/resources/news/t0-more-risk-fewer-benefits/.
Bodson, M. (2021). Written testimony of Michael Bodson, on behalf of Depository Trust and Clearing Corporation. Hearing before the US House of Representatives Committee on Financial Services, May 6. URL: http://www.congress.gov/117/meeting/house/112590/witnesses/HHRG-117-BA00-Wstate-BodsonM-20210506.pdf.
Byck, S., and Heijmans, R. (2020). How much liquidity would a liquidity-saving mechanism save if a liquidity-saving mechanism could save liquidity? A simulation approach for Canadaâs large-value payment system. Discussion Paper, February 18, Payments Canada. URL: http://www.payments.ca/insights/research/how-much-liquidity-would-liquidity-saving-mechanism-save-if-liquidity-saving.
Canadian Capital Markets Association (2023). One-year countdown until ${T+\text{1}}$ standard settlement goes live. Announcement, March 14, CCMA, Toronto. URL: https://bit.ly/3vHTWEK.
Clancy, L. (2022). Broken chains: how DLT code switch compounded ASX fail. Risk.net, December 8. URL: https://bit.ly/3tSQ57m.
CLS Group (2022). Annual report and consolidated accounts 2022. CLS Group Holdings AG, London. URL: http://www.cls-group.com/media/tjwjfcnz/cls_annual_report_2022.pdf.
Depository Trust and Clearing Corporation (2022). Accelerating to $T+\text{1}$ : what you need to know. Blog Post, June 6, DTCC Connection. URL: http://www.dtcc.com/dtcc-connection/articles/2022/june/06/accelerating-to-t1-what-you-need-to-know.
Duffie, D., and Zhu, H. (2011). Does a central clearing counterparty reduce counterparty risk? Review of Asset Pricing Studies 1(1), 74–95 (https://doi.org/10.1093/rapstu/rar001).
European Central Bank (2023). TARGET2-Securities Annual Report 2022. ECB, Frankfurt. URL: http://www.ecb.europa.eu/paym/intro/publications/html/ecb.t2sar2022.en.html.
European Securities and Markets Authority (2023). Call for evidence on shortening the settlement cycle. Discussion Paper, October 5, ESMA, Paris. URL: https://bit.ly/3SlsF49.
Feenan, S., Heller, D., Lipton, A., Morini, M., Ram, R., Sams, R., Swanson, T., Yong, S., and Zalles, D. B. (2021). Decentralized financial market infrastructures: evolution from intermediated structures to decentralized structures for financial agreements. Journal of FinTech 1(2), Paper 2150002 (https://doi.org/10.1142/S2705109921500024).
Global Financial Markets Association (2023). FX considerations for $T+\text{1}$ securities settlement. Report, May, GFMA Global FX Division, Washington, DC. URL: https://bit.ly/3O4hObY.
HM Treasury (2022). Accelerated Settlement Taskforce: terms of reference. Policy Paper, December 9, HM Treasury, London. URL: http://www.gov.uk/government/publications/accelerated-settlement-taskforce.
Kharif, O. (2023). JP Morgan is exploring blockchain-based deposit token for payments, settlements. Bloomberg, September 7. URL: http://www.bloomberg.com/news/articles/2023-09-07/jpmorgan-jpm-explores-blockchain-deposit-token-for-payment-settlement.
Mathew, G. (2023). New $T+\text{1}$ settlement cycle comes into effect today. Indian Express, January 27. URL: https://indianexpress.com/article/explained/explained-economics/t1-settlement-cycle-comes-into-effect-on-january-27-8406168/.
McLaughlin, D. (2024). Remarks on blockchain and distributed ledger technology in financial market infrastructures. In WFEClear: Proceedings of the 40th Annual Clearing and Derivatives Conference. World Federation of Exchanges, London. Forthcoming.
Monahon, K. (2019). Steampunk settlement: futuristic technology to achieve an anachronistic result. White Paper, June 11, Greenwich Associates. URL: http://www.greenwich.com/market-structure-technology/steampunk-settlement.
Pozmanter, M. (2021). Evolution to an accelerated settlement. Interview with DTCC Connection, June 21. URL: http://www.dtcc.com/dtcc-connection/articles/2021/june/21/evolution-to-an-accelerated-settlement-cycle.
US Securities and Exchange Commission (2022a). Shortening the securities transaction settlement cycle. Proposed Rule. Federal Register, 87 FR 10436. URL: http://www.sec.gov/files/rules/proposed/2022/34-94196.pdf.
US Securities and Exchange Commission (2022b). Comments on notice of shortening the securities transaction settlement cycle. Releases 34-94196, IA-5957; File S7-05-22. URL: http://www.sec.gov/comments/s7-05-22/s70522.htm.
US Securities and Exchange Commission (2023). Shortening the transactions settlement cycle. Final Rule. Code of Federal Regulations, 17 CFR 232, 240 and 275. URL: http://www.sec.gov/files/rules/final/2023/34-96930.pdf.

[bib.bib1] Asset Servicing Times (2021). $T+\text{2}=\text{antiquated}$ or reliable? Asset Servicing Times, March 3. URL: https://bit.ly/3vulNZh.

[bib.bib2] Bachmann Task Force (1992). Report of the Bachmann Task Force on clearance and settlement reform in US securities markets. Report, May, US Securities and Exchange Commission, Washington, DC. URL: https://bit.ly/3TVKXde.

[bib.bib3] Barabási, A.-L., and Pósfai, M. (2016). Network Science. Cambridge University Press.

[bib.bib4] Bentsen, K. E., Jr. (2022). $T+\text{0}$ ? More risk, fewer benefits. News Article, February 2, Securities Industry and Financial Markets Association, Washington, DC. URL: http://www.sifma.org/resources/news/t0-more-risk-fewer-benefits/.

[bib.bib5] Bodson, M. (2021). Written testimony of Michael Bodson, on behalf of Depository Trust and Clearing Corporation. Hearing before the US House of Representatives Committee on Financial Services, May 6. URL: http://www.congress.gov/117/meeting/house/112590/witnesses/HHRG-117-BA00-Wstate-BodsonM-20210506.pdf.

[bib.bib6] Byck, S., and Heijmans, R. (2020). How much liquidity would a liquidity-saving mechanism save if a liquidity-saving mechanism could save liquidity? A simulation approach for Canadaâs large-value payment system. Discussion Paper, February 18, Payments Canada. URL: http://www.payments.ca/insights/research/how-much-liquidity-would-liquidity-saving-mechanism-save-if-liquidity-saving.

[bib.bib7] Canadian Capital Markets Association (2023). One-year countdown until ${T+\text{1}}$ standard settlement goes live. Announcement, March 14, CCMA, Toronto. URL: https://bit.ly/3vHTWEK.

[bib.bib8] Clancy, L. (2022). Broken chains: how DLT code switch compounded ASX fail. Risk.net, December 8. URL: https://bit.ly/3tSQ57m.

[bib.bib9] CLS Group (2022). Annual report and consolidated accounts 2022. CLS Group Holdings AG, London. URL: http://www.cls-group.com/media/tjwjfcnz/cls_annual_report_2022.pdf.

[bib.bib10] Depository Trust and Clearing Corporation (2022). Accelerating to $T+\text{1}$ : what you need to know. Blog Post, June 6, DTCC Connection. URL: http://www.dtcc.com/dtcc-connection/articles/2022/june/06/accelerating-to-t1-what-you-need-to-know.

[bib.bib11] Duffie, D., and Zhu, H. (2011). Does a central clearing counterparty reduce counterparty risk? Review of Asset Pricing Studies 1(1), 74–95 (https://doi.org/10.1093/rapstu/rar001).

[bib.bib12] European Central Bank (2023). TARGET2-Securities Annual Report 2022. ECB, Frankfurt. URL: http://www.ecb.europa.eu/paym/intro/publications/html/ecb.t2sar2022.en.html.

[bib.bib13] European Securities and Markets Authority (2023). Call for evidence on shortening the settlement cycle. Discussion Paper, October 5, ESMA, Paris. URL: https://bit.ly/3SlsF49.

[bib.bib14] Feenan, S., Heller, D., Lipton, A., Morini, M., Ram, R., Sams, R., Swanson, T., Yong, S., and Zalles, D. B. (2021). Decentralized financial market infrastructures: evolution from intermediated structures to decentralized structures for financial agreements. Journal of FinTech 1(2), Paper 2150002 (https://doi.org/10.1142/S2705109921500024).

[bib.bib15] Global Financial Markets Association (2023). FX considerations for $T+\text{1}$ securities settlement. Report, May, GFMA Global FX Division, Washington, DC. URL: https://bit.ly/3O4hObY.

[bib.bib16] HM Treasury (2022). Accelerated Settlement Taskforce: terms of reference. Policy Paper, December 9, HM Treasury, London. URL: http://www.gov.uk/government/publications/accelerated-settlement-taskforce.

[bib.bib17] Kharif, O. (2023). JP Morgan is exploring blockchain-based deposit token for payments, settlements. Bloomberg, September 7. URL: http://www.bloomberg.com/news/articles/2023-09-07/jpmorgan-jpm-explores-blockchain-deposit-token-for-payment-settlement.

[bib.bib18] Mathew, G. (2023). New $T+\text{1}$ settlement cycle comes into effect today. Indian Express, January 27. URL: https://indianexpress.com/article/explained/explained-economics/t1-settlement-cycle-comes-into-effect-on-january-27-8406168/.

[bib.bib19] McLaughlin, D. (2024). Remarks on blockchain and distributed ledger technology in financial market infrastructures. In WFEClear: Proceedings of the 40th Annual Clearing and Derivatives Conference. World Federation of Exchanges, London. Forthcoming.

[bib.bib20] Monahon, K. (2019). Steampunk settlement: futuristic technology to achieve an anachronistic result. White Paper, June 11, Greenwich Associates. URL: http://www.greenwich.com/market-structure-technology/steampunk-settlement.

[bib.bib21] Pozmanter, M. (2021). Evolution to an accelerated settlement. Interview with DTCC Connection, June 21. URL: http://www.dtcc.com/dtcc-connection/articles/2021/june/21/evolution-to-an-accelerated-settlement-cycle.

[bib.bib22] US Securities and Exchange Commission (2022a). Shortening the securities transaction settlement cycle. Proposed Rule. Federal Register, 87 FR 10436. URL: http://www.sec.gov/files/rules/proposed/2022/34-94196.pdf.

[bib.bib23] US Securities and Exchange Commission (2022b). Comments on notice of shortening the securities transaction settlement cycle. Releases 34-94196, IA-5957; File S7-05-22. URL: http://www.sec.gov/comments/s7-05-22/s70522.htm.

[bib.bib24] US Securities and Exchange Commission (2023). Shortening the transactions settlement cycle. Final Rule. Code of Federal Regulations, 17 CFR 232, 240 and 275. URL: http://www.sec.gov/files/rules/final/2023/34-96930.pdf.

Journal of Financial Market Infrastructures