Optimal Settlement Speed

“In the world of one-click purchasing, instant payments, and two-hour delivery, many investors find it counterintuitive that securities trades in the United States can take two days to complete or settle.”
Michael Bodson, President and CEO of DTCC, Business Insider, April 2, 2021.

Over the past year, a series of events has shifted the attention of both experts and laypersons to the arcane processes that support trading and settlement in the U.S. securities markets. The massive volume of U.S. Treasury sales in March 2020 at the start of the COVID crisis boosted liquidity needs at precisely the time when resources were scarce, overwhelming the over-the-counter trading system and compelling unprecedented Fed interventions (see our earlier post). Similarly, the late-January 2021 episode involving extraordinary activity in GameStop, in large part through Robinhood (the broker-dealer), highlighted how a surge in equities trading volume can concentrate counterparty risk and trigger a jump in liquidity needs to settle those trades (see our earlier post). After filling the news for weeks, the equity market turmoil triggered Congressional hearings (see here and here).

In an effort to reduce both counterparty risk and liquidity needs, a number of observers are focusing on shortening the settlement period. Officials at the Depository Trust and Clearing Corporation (DTCC), whose subsidiaries in 2020 cleared a daily average of more than $9 trillion in a range of instruments, say they are prepared to halve the equities settlement period from two days (T+2) to one (T+1), calling on the industry to do so by 2023. Others have called for far faster clearing, including nearly instant settlement (see here).

Times like these lead us to ask: how can we improve the efficiency and safety of the financial plumbing? We see plenty of room for progress in speeding settlement, thereby reducing both risk and liquidity needs during periods of stress. However, we also think that things can go too far (or too fast). In particular, we are deeply skeptical of calls for real-time settlement (T+0) for securities or foreign exchange transactions. In this post, we suggest why the optimal settlement period for some financial transactions is not zero.

Let’s start with payments for goods and services, where the case for real-time gross settlement (RTGS) is well-known: vendors seek finality, not counterparty risk, when they make a sale. A retailer rightly expects payment before a customer can walk out with the merchandise. Grocers, restaurants, car dealers, and the like are not interested in the risk that would arise if they sold everything on credit and then had to wait for payments to arrive.

As a result, many advanced economy central banks support RTGS for both retail payments and transfers among financial intermediaries (including the United Kingdom, which introduced Faster Payments in 2008). Importantly, RTGS systems typically are built around the confidence of intermediaries in the ability of a monopolist public institution—typically a central bank—to ensure timely performance.

In the United States, retail vendors already have payments finality through credit and debit transactions supported by private networks like VISA and Mastercard. However, transfers among intermediaries often involve considerable delays in granting account holders access to funds. The Federal Reserve is working to improve the speed and efficiency of the U.S. payments system through its FedNow service, which it aims to make operational by 2023. However, the Fed also lags in broadening its wholesale payments operations.

In the case of securities, there has been a centuries-long march toward faster settlement. Already in the 1700s, the integration of the Amsterdam and London stock markets allowed for cross-border settlement in 14 days (T+14)—the time required for delivering the exchanged (paper) instruments. In the U.S. equity market, the settlement period shortened from T+5 days in 1975 (set by the SEC) to T+3 in 1995 and T+2 in 2017. With technology available today, there is considerable scope for faster settlement for a wide range of financial instruments, including U.S. stocks and corporate and municipal bonds, which are all still T+2; as well as Treasury instruments and foreign exchange, which are mostly T+1.

By reducing costly collateral requirements (i.e., liquidity needs), shortening these settlement periods would improve efficiency without sacrificing safety. These requirements typically reflect uncertainty about the ability of a counterparty to pay in the event of a large price swing during the settlement period. Thus, DTCC estimates that a shift from T+2 to T+1 settlement in the equity market would lower counterparty risks sufficiently to reduce this “volatility component” of collateral requirements by 41 percent (see here, page 5). Presumably, what we think of as “T+0 end-of-day” settlement—where all the trades for a given day are final at the end of the trading day—would lower collateral needs still further.

However, we doubt that going all the way to real-time settlement (truly T+0) would be appropriate even as an aspirational goal for securities or foreign exchange transactions. We say this for several reasons. First, the gross volume of purely financial transactions vastly exceeds the volume of retail payments. Consider, for example, only the average daily gross transactions in foreign exchange involving the U.S. dollar (about $5.8 trillion; see Table 1), in the U.S. fixed-income market (about $1 trillion; see here) and in the U.S. equities market (about $0.5 trillion; see here). At a combined $7.3 trillion, these are nearly 500 times business-daily U.S. retail sales (about $15 billion)! And that leaves aside the vast market for dollar-denominated derivative instruments.

Second, with a long enough settlement period, central clearing parties (CCPs) can substantially reduce the volume of payments through what is known as multilateral trade compression or netting. To see how the process works, assume that firm A sells a security to firm B, then firm B sells it to firm C, and finally, firm C sells it back to firm A. If all these trades are within the settlement period, a CCP can net them so that there are no payments at all (see our earlier post)! The length of time needed for efficient netting depends on the technology not only of the CCP, but also of its clearing members (CMs). For example, DTCC has explored “T+0 end of day” settlement with market participants, but concluded in its T+1 proposal for 2023 that the industry currently is not ready for same-day settlement (see page 7 here).

Shortening the settlement period to T+0—real time gross settlement of each and every trade—eliminates the possibility of netting, making it fundamentally different from “T+0 end of day” settlement. In a T+0 regime, all trades would become “delivery versus payment,” massively increasing both liquidity and security inventory needs, as well as the risk that the transactions would fail to clear (what are known as “fails”). Indeed, instant settlement would require that sellers trust their anonymous counterparts to make payments on time and in full, rather than relying on the CCPs, which are designed to reduce such counterparty risk.

The leading CCPs—such as DTCC’s subsidiaries and the Chicago Mercantile Exchange (CME)—often serve as virtual monopolies. Their official designation as financial market utilities (FMUs) provides them with access to Fed liquidity (assuming they remain solvent) and highlights that they also are too big to fail. Everyone in the financial system has powerful incentives to ensure continuity of business were a CCP to become insolvent, so it is no surprise that they dominate financial settlement today.

In addition to promoting CCP resilience, the concern of market participants and regulators alike should be one common to monopolists: to give CCPs adequate incentives to innovate and lower industry-wide costs. Faced with a coordination problem among their CMs, CCPs also need regulatory support to motivate laggard CMs who may be reluctant to make the investments necessary for secure, faster settlement.

To get a sense of the huge liquidity benefits of netting and why a move to real-time settlement for securities transactions would be a mistake, consider a few examples. DTCC estimates that netting reduces payments to settle U.S. equity transactions by an average of 98 percent each day! To see the point, consider the case of March 3, 2020—a heavy trading day at the beginning of the pandemic. That day, gross U.S. equity transactions were $3.5 trillion. After netting, final settlement required payments of a mere $80 billion. Similarly, in the foreign exchange market, the leading CCP (CLS Bank) reports (in its latest disclosure) that a combination of netting and other liquidity management tools reduced 2019 average daily payments from a gross level of $5.9 trillion to about $150 billion.

In the U.S. Treasury market, central clearing of inter-dealer trades already reduces payment needs substantially. Keane and Fleming use nonpublic data on Treasury transactions to estimate the further benefits that would arise if dealer-to-customer Treasury trades also were centrally cleared. The following chart (based on their Table 1) shows a reduction of average daily liquidity needs in the January-April 2020 period by 61 percent, from $598 billion to $232 billion. Importantly, the additional netting benefits exceed this norm on high-volume trading days (like February 28 or March 30 of 2020), when liquidity needs are greatest.

Treasury Dealers: Reduction of Gross Settlement Obligations Through Central Clearing (Jan-Apr 2020)

Note: The totals that appear above the bars already incorporate the impact of netting dealer-to-dealer trades. The red bars show the estimated impact of imposing central clearing on dealer-to-customer trades. Source: Table 1, Fleming and Keane, The Netting Efficiencies of Marketwide Central Clearing.

If anything, these sizable estimates understate the benefits from mandating central clearing for Treasuries (see our earlier post). First, as Keane and Fleming note, they do not include the potential further reduction that could be gained from central clearing of dealers’ purchases in Treasury auctions and of their repo transactions. Second, the authors show that most fails among uncleared Treasury transactions arise from “daisy-chain” events in which an initial failure to deliver triggers another (and so on, and so forth). By interrupting these chains, central clearing for all trades could slash fails (and the costly dislocations they cause) down to the small gray shaded area (labeled “Remaining”) in the following chart. Again, the biggest reductions occur in periods of financial stress when trading activity tends to be highest.

Treasury Primary Dealers’ Matched and Remaining Fails in Specific Issues, Jan-Apr 2020

Source: Figure 13, Fleming and Keane, The Netting Efficiencies of Marketwide Central Clearing.

So, where does all this leave us? We strongly share the desire for secure, faster, and stable settlement systems. In the United States, we hope that the Fed speeds the long-overdue introduction of U.S. real-time payments (24x7x365). As for securities and foreign exchange markets, we hope that regulators and CCPs will aim for “T+0 end-of-day” settlement as soon as possible, while preserving the benefits of multilateral netting and reliable execution.

To make secure and faster settlement a practical (rather than an aspirational) goal, regulators may need to monitor progress (and even set timetables) in ways that provide incentives to those involved. Both CCPs and (perhaps more importantly) their CMs will need to make investments today to realize the aggregate savings from reducing liquidity needs (and fails) tomorrow. While these future benefits will accrue gradually over time, they will be especially important when the financial system again comes under severe stress. Given the spillover effects across markets of changing settlement arrangements, combined with the fragmented U.S. regulatory framework, the Financial Stability Oversight Council appears to have a natural role in helping to harmonize these regulatory efforts.

The good news is that—based on experience with previous moves to speed settlement—coordinating such industry-wide technical progress is feasible. At the same time, we should be wary of trying to go either too fast or too far. Going too fast risks making market disruptions both more frequent and more severe, while going too far risks sacrificing the enormous benefits we get from central clearing.

Acknowledgements: Without implicating him, we thank Dino Kos for guiding us to useful sources regarding foreign exchange clearing and settlement. We also thank Michael Fleming and Frank Keane for the very helpful materials in their recent paper, The Netting Efficiencies of Marketwide Central Clearing.