Left feeling unsettled: what are settlement failures, how prevalent are they, and what do we do about them?

Gary Harper, Pedro Gurrola-Perez and Jieshuang He

What is a settlement fail?

Imagine you’ve booked tickets for a flight, and go to pick them up and pay for them on the day. You arrive at the airport but find out the airline has overbooked, and already given your ticket away. Worse yet, because you’ve missed this flight you’re going to miss an onward connection. But, you’ll likely get a replacement ticket in a day or two as compensation.

Now imagine you’re a bank (actually, you could be a bank, investment firm, insurer, or pension fund – but for this post we’re just using banks as a catch-all), the plane ticket is £10 million worth of securities, and your onward connection is actually another bank you’re meant to pass the securities on to. This is called a settlement fail and it’s a more common occurrence than you’d think, importantly, the market understands that a settlement fail is usually temporary and doesn’t constitute a default.

Settlement fails aren’t limited to Bank A not being able to pass on a security to Bank B. Bank B may need to collateralise a repo with investment firm C using the security. And investment firm C may have a margin payment to make using the security it got from Bank B. Security fails can propagate through a number of different counterparties producing a cascade’ effect. Cascades can lead to closed loops, where Bank A might be waiting to receive the security from investment firm C. In this case a gridlock happens and no one can deliver securities to their counterparties.

Fails can have two knock on effects for market functioning. First, the gridlock effect could impair functioning as very few securities may end up being delivered in such a gridlock. Second, lenders may be reluctant to lend out their securities if they think they won’t get them back. So we published a Staff Working Paper that looks at fails in the UK gilts and equity (of companies listed on the FTSE-100) markets, the paper considers the following questions:

A) How often do fails happen?

B) What causes them?

C) How can we tell if a fail has happened because of a cascade effect?

D) Is there an efficient way of dealing with cascades of fails when they happen?

Our conclusions on these topics are based on data we received from Euroclear UK & Ireland (the UK’s Central Securities Depository), which operates CREST (the UK’s securities settlement system). The data cover fails in the FTSE-100 and gilt markets from 2016 Q4 to 2017 Q1. We looked at these two markets since they’re some of the largest and most systemically important markets in the UK.

How frequent are fails?

More often than you might think, as shown below.

Chart 1: Daily fail ratio of gilts

Chart 2: Daily fail ratio of equities

We don’t see any particular trends in fails for either market and at a minimum we found that around 2%–10% of gilt trades and 4%–16% of FTSE-100 trades by volume failed on a given day.

We also found that it’s not just a few firms that would sometimes fail to settle. There’s a pretty wide distribution of average fail rates in both markets as shown below. Some banks tended to fail a very high proportion of their trades, but that’s because they were carrying out very low volumes of trades.

Chart 3: Distribution of average fail ratios for gilts

Chart 4: Distribution of average fail ratios for FTSE-100 equities

Finally, fails can last for more than one day. Much like the airline giving you a ticket for a flight the next day, banks can hand over securities after the date they should have settled. We found in both markets between 30%–40% of a given days’ fails had actually persisted for more than one day.

What causes them?

There are three main reasons for failing to deliver a security:

1) Operational problems – the banks’ back office may not be able to provide securities to settle trades struck by its front office.

2) Liquidity problems – the firm can’t get hold of the security either because of high demand or because it was relying on getting the security from a counterparty (Cascade of fails).

3) Strategic behaviour – The firm might have got a better offer for a security or it may cost less to fail than it does to borrow a security to honour its obligations.

We were mostly looking at whether operational failures and strategic behaviour were behind banks failing. So we looked at the gilt/FTSE-100 markets since they are the most liquid settled in CREST and liquidity problems should be minimal in the markets.

To try and spot if operational problems were causing fails we looked at whether particular banks tended to fail particular securities (we distinguish securities using their International Securities Identification Number, or ISIN). If specific ISINs tended to fail more frequently, we’d see a line extending from the ISIN axis up near the top of the 3d charts. If a specific bank tended to fail irrespective of ISIN, then we’d see a line extending from the seller axis high up in the charts. But neither of these lines are in the charts, so we didn’t think that particular banks or particular ISINs tended to fail more often than others.

Chart 5: Proportion of days where a seller failed for a particular gilt ISIN

Chart 6: Proportion of days where a seller failed for a particular FTSE-100 ISIN

We did, however find that banks that either:

A) Traded a wider variety of securities.

B) Traded a low volume of securities.

Tended on average to have higher fail ratios, with the low volume banks being the ones that tended to have the very high fail ratios (80%–90%). We think that banks failing on a bigger range of ISINs suggests they are failing due to operational reasons. If, instead, banks were acting strategically, they would typically fail to settle trades involving ISINs that have recently spiked in value, as they can negotiate a higher price for them than already agreed with their current counterparty. This would mean fails would be concentrated in these ISINs – which is not what we see in the data. 

How to tell if a fail happened because of a cascade effect

As we know, some fails happen because the firm was waiting to receive the security from a counterparty. We wanted to understand how cascades of fails affect the network of settlement fails. Using network analysis we developed an algorithm to try and identify whether a specific fail has been caused by a cascade effect or not, we found that around 60% of sellers fail at least some of their trades due to cascade effects on a given day.

We needed to find cases where a bank had failed to settle trades involving a high value of securities, but were only expecting to receive a low value of the same securities from their counterparties. If the bank was expecting to receive more securities than it failed to deliver, then in all likelihood, it only failed because of its counterparties and was caught in a cascade. Figure 1 below shows how this works in principle, it shows three banks trading different values of the same ISIN and identifies which fails have occurred due to cascade effects.

Figure 1: Finding fails that occurred due to cascade effects

How can we deal with cascades of fails?

Since the effects of fails can be significant, many jurisdictions have come up with ways of resolving them. The two main methods being to either use fines for fails or a process known as a buy-in’.

A buy-in means that the bank that was due to receive securities employs a third party to supply the securities. This third party then bills the original securities seller for the cost of the securities. Extending the airline ticket example, a buy-in is as if you’ve bought a replacement ticket from another airline and then got the original airline to pay for it.

Buy-ins currently work on a bilateral basis, so every bank buys in their counterparty and passes securities down a chain. But if there’s a network of settlement fails, such as in a cascade, this could result in a lot of banks buying each other in for the same security.

So we looked into the network of fails in the markets we analysed and found we could adapt the framework from Eisenberg and Noe (which is used to analyse how to clear obligations in payment systems, especially in the instance that one member has defaulted) to create a centralised buy-in strategy. We found it is possible to resolve networks of fails by buying in only those banks that are net sellers of the security. This is illustrated below in Figure 2 (showing the network of fails for a random gilt ISIN on a random day) where the network is resolved by buying in the nodes coloured in yellow, the blue nodes will simply pass on the securities they owe when they receive them from their counterparty.

Figure 2: Example of a centralised buy-in strategy for a random gilt ISIN

Limiting settlement fails and remedying them quickly when they happen is important for market functioning. Our paper looks at a centralised strategy for carrying out buy-ins as one way of sorting out gridlocks of settlement fails. We don’t think this strategy has been tried before though, so there could be a lot of operational issues with actually implementing it.

We acknowledge and thank Euroclear UK & Ireland for providing us with the data.

Gary Harper works in the Bank’s Stress Testing Strategy Division, Pedro Gurrola-Perez formerly worked in the Bank’s Financial Market Infrastructure Directorate and Jieshuang He works at the Chinese University of Hong Kong, Shenzen, China.

If you want to get in touch, please email us at bankunderground@bankofengland.co.uk or leave a comment below.

Comments will only appear once approved by a moderator, and are only published where a full name is supplied. Bank Underground is a blog for Bank of England staff to share views that challenge – or support – prevailing policy orthodoxies. The views expressed here are those of the authors, and are not necessarily those of the Bank of England, or its policy committees.