Quants don’t typically spend a lot of time poring over legal and regulatory documents. That’s a shame because, when they do, the results can be spectacular.
A rare example this year came in the field of settlement risk – the danger that one party to a trade makes a trade or variation margin payment, while the other does not. In the presence of initial margin, which is collected as an additional buffer against a party’s default, the assumption might be that this risk is eliminated. By picking apart the mechanics of industry standard margining contracts, however, a trio of quants revealed just how much exposure remains – a staggering five to ten times higher than previously thought.
“They looked at the entire complexities of the margining process and modelled it mathematically,” says Alexei Kondratyev, a managing director at Standard Chartered in London. “They looked at things from first principles and the result was amazing.”
In Does initial margin eliminate counterparty risk?, Leif Andersen, global co-head of the quantitative strategies group at Bank of America Merrill Lynch, Michael Pykhtin, a manager in the quantitative risk team at the Federal Reserve Board in Washington, DC, and Alexander Sokol, chief executive officer of technology vendor CompatibL, built a model to quantify settlement risk and showed it can create massive exposures during certain periods in the margining cycle – even in the presence of initial margin.
In addition, they suggested solutions that are now being explored by market participants.
Where does the exposure come from? The research highlighted two main sources. First, under existing variation margining practices in the OTC market, there is usually a time lag between trade cashflows and margin payments.
Typically, the market data and trade data as at end of day T is used to value the amount of collateral to be posted the next day, or T+1. Once this has been instructed, the bank would typically receive the margin on T+2. This means there is always a gap between the trade payments and the margin payments, creating ‘spikes’ in the exposure to the counterparty.
This is a particular problem when there is a large trade cashflow during any given period that would then significantly move the mark-to-market on that trade – but the margin is not due until days later.
A second source is rooted in human behaviour. On the brink of default, a counterparty will sometimes dispute a margin call to buy time. During this period, the bank is expected to continue honouring trade cashflows, as failing to do so may be considered a default. This could extend the margin period of risk, or the gap between the date of the last valuation on which the margin was paid in full and the close-out date, allowing exposures to build.
We deal with many banks, we see what people do for XVAs, and I saw that every single bank treats it slightly differentlyAlexander Sokol, CompatibL
A convenient shortcut some risk managers have been known to use to deal with the spikes for pricing and capital calculations is removing them altogether from the exposure profile. Some banks may also take a more proactive approach by deciding when they will stop paying either a margin flow or a trade flow, such as a swap payment, for instance.
But Andersen, Pykhtin and Sokol didn’t want to settle for shortcuts.
The three started working together on the paper in 2014, when independent research they carried out within their own institutions led them to question existing market practice around how cashflows and margin payments are modelled .
For CompatibL’s Sokol, that work was prompted by knowledge of how his clients calculate counterparty credit exposures and the accompanying family of valuation adjustments, known as XVAs.
“We deal with many banks, we see what people do for XVAs, and I saw that every single bank treats it slightly differently. Some people were not expecting any cashflows during the margin period of risk, other people said they were expecting to pay and receive cashflows of the trade but not the margin, whereas some people were expecting only to pay them,” says Sokol. “It’s almost random. It depends on who the quant is at the bank to make that choice, but these numbers end up on the financials of the bank – it is material information for shareholders. So it seemed like it was the right time to figure out the right approach.”
For BAML’s Andersen and the Federal Reserve’s Pykhtin, both now two-time winners of the award, interest was sparked when working on various regulatory initiatives.
Those banks that feel they can sniff out default risk a mile away should feel free to be more aggressive on the parameterisation and assume a shorter margin period of riskLeif Andersen, Bank of America Merrill Lynch
For Pykhtin, counterparty credit risk is a key area of research as part of his role at the Fed. He was awarded the title of quant of the year in 2014 for his work on systemic risk – a paper co-authored by Sokol – and counterparty exposure modelling.
BAML’s Andersen started looking at the issue while working with the industry committee that was developing a standard model to calculate initial margin for non-cleared derivatives.
The work required a deep understanding of what contributes to counterparty exposure. It was during this time that Andersen learnt the workings of the Isda CSA by talking to lawyers and collateral managers. Eventually, he produced what he calls a “quant-friendly” and detailed description of the margining process and the steps taken in case of missed payments.
“Basically Leif was talking to people who do it every day to understand exactly how it works. When somebody defaults, who sends the letter? How long does it take to draft a letter? How many days are there in a grace period?” says CompatibL’s Sokol. “There was analysis of unprecedented detail, and then we spent a few months until it crystallised exactly how to capture it into a model.”
The result was a four-parameter model that could not only quantify settlement risk, but also capture behavioural elements introduced by the contractual and legal aspects of a CSA – the possibility that a margin dispute might be deliberately drawn out to postpone an imminent default, for example.
As things stand, most banks calculate their counterparty exposure using a single parameter – the length of the margin period of risk, which is assumed to be a fixed 10 business days. The quants’ four parameters, on the other hand, capture different stages of the margin period of risk – depending on different contractual obligations and scenarios. The end for all four periods is the date on which the portfolio is finally closed out. The starting points are the calculation day of the defaulting party’s last full margin payment – and the same for its counterparty – plus the time at which the defaulting party stopped paying trade cashflows, and the corresponding date for the second party.
This kind of detailed parameterisation allows for flexibility within the model and also a way to align it with the views of the legal and operations departments of the bank. For instance, some banks in the past have withheld trade cashflows in instances where they had reason to believe the counterparty was distressed, the legal department had approved the decision.
“They can customise the model parameters the way they want. Those banks that feel they can sniff out default risk a mile away should feel free to be more aggressive on the parameterisation and assume a shorter margin period of risk … and you can justify to internal and external governance processes how you go about it,” says Andersen.
Others agree the flexibility has benefits.
“Their paper changed the way we should think about exposures. Instead of calculating exposures based on a 10-day margin period of risk, which is the standard, they looked at the mechanics of margin, such as who pays whom and under which situation. And that’s a more accurate way of looking at it,” says Standard Chartered’s Kondratyev.
People used to do return on assets under management, but return on cash is becoming more [common], because a lot of funds are being redeemed… so having something that can reduce your margin is very usefulSenior executive at a US-based hedge fund
Given this kind of detail, the paper has already stirred an industry-wide dialogue on how to address settlement risk.
A solution originally formulated by CompatibL’s Sokol, which was later included in the paper, was to calculate margin calls based on portfolio value as of two days later instead of a given day, so the margin payment – which was scheduled to arrive two days after the margin call – arrives on the same day as the trade flow, and the two can be netted, effectively closing the margin gap.
According to the quants, the risk can be further reduced by settling the netted payment through CLS Bank, which uses a payment-versus-payment system for foreign exchange settlement – that is, one party gets paid only when the other pays.
Some buy-side firms are already considering the first solution, and are starting to have discussions with banks on how they could amend their CSAs so they can start netting payments.
“People used to do return on assets under management, but return on cash is becoming more [common], because a lot of funds are being redeemed… so having something that can reduce your margin is very useful,” says one senior executive at a US-based hedge fund.
Regulators have taken note too. The European Central Bank’s guidance on its review of internal models, launched in 2015, includes a section on the treatment of margin and trade payments within the margin period of risk that is said to have been influenced by the quants’ research.
As for the future, the quants say they will continue exploring topics that can have an industry-wide impact. Fintech is one such topic for BAML’s Andersen and CompatibL’s Sokol.
“Fintech is a big thing… and goes hand-in-hand with a variety of activities around data gathering, algorithmic trading and electronic trading, areas a lot of banks are interested in, and areas regulators encourage through clearing and transparency requirements and so forth,” says Andersen.
The shift towards activities that require more efficiency means working on ways to optimise time and costs will be a key focus for the quants.
“It’s not about pricing anymore. It’s about spreading analytics to every nook and cranny of the bank,” adds Andersen.
Sokol says his firm will continue focusing on offering trading and risk analytics to clients, with quantitative research and innovation remaining a priority within the firm.